FUNDAMENTAL STUDIES OF CATALYTIC DEHYDROGENATION ON
ALUMINA-SUPPORTED SIZE-SELECTED PLATINUM CLUSTER
MODEL CATALYSTS
by
Eric Thomas Baxter
A dissertation submitted to the faculty of The University of Utah in partial fulfillment of the requirements for the degree of
Doctor of Philosophy
Department of Chemistry
The University of Utah
May 2018
Copyright © Eric Thomas Baxter 2018
All Rights Reserved
T h e U n iversity of Utah Graduate School
STATEMENT OF DISSERTATION APPROVAL
The dissertation of Eric Thomas Baxter has been approved by the following supervisory committee members:
Scott L. Anderson , Chair 11/13/2017 Date Approved
Peter B. Armentrout , Member 11/13/2017 Date Approved
Marc D. Porter , Member 11/13/2017 Date Approved
Ilya Zharov , Member 11/13/2017 Date Approved
Sivaraman Guruswamy , Member 11/13/2017 Date Approved
and by Cynthia J. Burrows , Chair/Dean of the Department/College/School of Chemistry and by David B. Kieda, Dean of The Graduate School.
ABSTRACT
The research presented in this dissertation focuses on the use of platinum-based catalysts to enhance endothermic fuel cooling. Chapter 1 gives a brief introduction to the motivation for this work. Chapter 2 presents fundamental studies on the catalytic dehydrogenation of ethylene by size-selected Ptn (n = 4, 7, 8) clusters deposited onto thin film alumina supports. The model catalysts were probed by a combination of experimental and theoretical techniques including; temperature-programmed desorption and reaction (TPD/R), low energy ion scattering spectroscopy (ISS), X-ray photoelectron spectroscopy (XPS), plane wave density-functional theory (PW-DFT), and statistical mechanical theory. It is shown that the Pt clusters dehydrogenated approximately half of the initially adsorbed ethylene, leading to deactivation of the catalyst via (coking) carbon deposition. The catalytic activity was observed to be size-dependent and strongly correlated to the cluster structure, with Pt7 demonstrating the highest activity.
In Chapter 3 the focus turns to selectively doping Pt7 clusters with boron. A combination of experiment and theory were used investigate the alkene-binding affinity of the bimetallic (PtnBm/alumina) model catalysts. A comparison of the theoretical and experimental results show that doping the Pt clusters with boron modifies the alkene- binding affinity and thus the tendency toward dehydrogenation to coke precursors.
Chapter 4 describes a way to produce bimetallic (PtnBm/alumina) model catalysts by exposing prepared Ptn/alumina samples to diborane and heating. It is shown that the diborane exposure/hearting procedure results in the preferential binding of B to the Pt clusters.
iv TABLE OF CONTENTS
ABSTRACT ...... iii
ACKNOWLEDGMENTS ...... vii
Chapters
1. INTRODUCTION ...... 1
1.1 Motivation ...... 2 1.2 References ...... 4
2. ETHYLENE DEHYDROGENATION ON Pt4, 7, 8 CLUSTERS ON Al2O3: STRONG CLUSTER SIZE DEPENDENCE LINKED TO PREFERRED CATALYST MORPHOLOGIES ...... 6
2.1 Overview ...... 7 2.2 Introduction ...... 7 2.3 Results and Discussion ...... 9 2.3.1 Cluster Catalyst Structures ...... 9 2.3.2 Size-Dependent Catalytic Activity ...... 11 2.3.3 Routes of Deactivation ...... 22 2.4 Conclusions ...... 31 2.5 Methodology ...... 32 2.5.1 Experimental Section ...... 32 2.5.2 Computational ...... 37 2.6 Acknowledgments...... 39 2.7 References ...... 39
3. BORON SWITCH FOR SELECTIVITY OF CATALYTIC DEHYDROGENATION ON SIZE-SELECTED Pt CLUSTERS ON Al2O3 ...... 53
3.1 Overview ...... 54 3.2 Introduction ...... 54 3.3 Results and Discussion ...... 56 3.4 Conclusion ...... 68 3.5 Methods...... 68 3.5.1 Experimental Section ...... 68 3.5.2 Computational ...... 70 3.6 Acknowledgements ...... 71 3.7 References ...... 72 4. DIBORANE INTERACTIONS WITH Pt7/ALUMINA: PREPARATION OF SIZE-CONTROLLED BORATED Pt MODEL CATALYSTS WITH IMPROVED COKING RESISTANCE ...... 80
4.1 Overview ...... 81 4.2 Introduction ...... 81 4.3 Methods...... 83 4.3.1 Computational ...... 84 4.3.2 Experimental ...... 85 4.4 Results ...... 89 4.4.1 Temperature Programmed Desorption/Reaction Following Adsorption of B2H6 and D2 ...... 89 4.4.2 X-Ray Photoelectron Spectroscopy ...... 93 4.4.3 Temperature-Dependent Ion Scattering Spectroscopy ...... 95 4.4.4 DFT Results for Adsorption of Diborane on Pt7 Clusters...... 97 4.4.5 Molecular Dynamics Simulations of Diborane/Pt7/Alumina Thermal Chemistry ...... 99 4.4.6 Pt4B4/Alumina ...... 102 4.5 Discussion ...... 104 4.5.1 Decomposition of Diborane on Pt7/Alumina ...... 104 4.6 Conclusion ...... 113 4.7 Acknowledgement ...... 113 4.8 References ...... 114
5. CONCLUSION ...... 127
Appendices
A: SUPPORTING INFORMATION FOR CHAPTER 2 ...... 130
B: SUPPORTING INFORMATION FOR CHAPTER 3 ...... 155
C: SUPPORTING INFORMATION FOR CHAPTER 4 ...... 177
vi ACKNOWLEDGMENTS
I would like to thank Scott L. Anderson for giving me the opportunity to work in his laboratory, first as an undergraduate student, and then as a graduate student. His willingness to have discussions, share his experiences, and to provide encouragement has played a vital role in my development as a research scientist.
I would also like to thank the other members of my committee: Professor Peter B.
Armentrout, Professor Marc D. Porter, Professor Ilya Zharov, and Professor Sivaraman
Guruswamy for constantly challenging me as a scientist.
I would also like to thank the members of the Anderson group for creating an enjoyable environment to learn and develop as a scientist.
Most of all, I would like to thank my family: my parents Jeffrey and Tonya, my sister, Stephanie, and my brother, Matthew. Without their endless love and support I would never have had the drive to make it this far in my academic career. CHAPTER 1
INTRODUCTION 2
1.1 Motivation
Hypersonic flight could revolutionize both commercial and commuter travel by significantly reducing the travel time. For example, travel from London to New York would take approximately two hours at hypersonic speeds. However, hypersonic flight has a major problem when it comes to dealing with the increased vehicle and engine heat as the aircraft speeds are increased. To sustain hypersonic flight would require a heat sink of approximately 3,500 kJ/kg to cool the engine and other vehicle components when traveling at speeds eight times the speed of sound.1-5 The fuel may be the most promising source of cooling. The required heat sink of ~3,500 kJ/kg, equivalent to temperatures on the order of 700°C, can be achieved by flowing fuels that undergo endothermic (heat- adsorbing) reactions through channels machined into the walls of the vehicle, engine, and subsystems to adsorb the waste heat. Unfortunately, solid carbon deposits begin to form as the temperature of the fuel rises beyond the onset of thermal cracking (~750 K) for hydrocarbons. Solid carbon deposits (“coke”) can interfere with heat transfer, clog fuel lines and filters, and disable fuel controls eventually leading to a system failure. Coke formation can occur through several mechanisms: interactions between hydrocarbons and the metal surfaces results in the formation of filamentous carbon; deposition of high- molecular-weight liquids (tar) results in amorphous carbon; amorphous carbon can also be formed on the surface from addition of small molecules and radicals. The formation of filamentous carbon can be avoided by careful selection of surface materials. The use of catalysts to selectively drive desired endothermic reactions (dehydrogenation of an alkane to produce an alkene and hydrogen) at temperatures below (~750 K) the onset of thermal cracking could provide a way to circumvent the formation of amorphous carbon.5 3
Platinum’s ability to (de)hydrogenate hydrocarbons has been extensively studied making it a material of interest in many catalytic applications.6,7 However, platinum catalysts have a tendency to deactivate during dehydrogenation reactions due to an accumulation of coke precursors. DFT-calculated potential energy surfaces for ethane dehydrogenation on the step and terrace sites of a Pt(433) single crystal show that ethane can readily dehydrogenate to ethylene.8 However, on the under-coordinated sites (e.g., step sites) further dehydrogenation is enhanced leading to the formation of coke precursors and the deactivation of the catalysts. One strategy for minimizing the formation of coke might be to selectively block the highly active under-coordinated sites.
It has been demonstrated that boron-doping the extended surfaces of Co9 and Ni10 extends the lifetime of the catalyst without compromising their catalytic activity in Fisher
Tropsch synthesis and steam reforming, respectively. The results of the boron-doping were attributed to boron selectively blocking the step and subsurface octahedral sites, therefore, reducing the nucleation of graphene islands.
Model catalysts with atomically size-selected metal clusters deposited on well characterized supports can provide a useful means for studying catalytic mechanisms by allowing for independent control of the size and density of active catalytic sites.11-19
Metal alloys/bimetallic catalysts can provide a valuable opportunity to tune catalytic activity, selectivity, and stability. However, extending the size-selected metal cluster approach to bimetallic clusters can be difficult. Synthesis of bimetallic clusters generally results in a wide distribution of cluster composition and structures limiting the ability to derive structure-property relations using characterization techniques (e.g., x-ray photoelectron spectroscopy) that provide information averaged over all the sites. 4
The work included within this dissertation provides results from studying the interactions between ethylene and alumina supported size-selected Ptn (n= 4, 7, 8) model catalysts with and without boron doping. The results show that Pt7 both binds and dehydrogenates ethylene more efficiently than Pt4 and Pt8, which is correlated to the diverse distribution of cluster morphologies accessible to Pt7 under the reaction conditions. Exposing the Pt7/alumina samples to diborane followed by heating results in the preferential deposition of boron atoms onto the Pt clusters. Boration of Pt clusters reduces the ethylene binding energies, such that ethylene preferentially desorbs intact rather than dehydrogenating to form coke precursors.
1.2 References
1. Lander, H. R., Jr.; Nixon, A. C. Endothermic Fuels for Hypersonic Vehicles. J. Aircr. 1971, 8, 200- 7.
2. Maurice, L.; Edwards, T.; Griffiths, J. Liquid Hydrocarbon Fuels for Hypersonic Propulsion. Prog. Astronaut. Aeronaut. 2000, 189, 757-822.
3. Edwards, T. Liquid Fuels and Propellants for Aerospace Propulsion: 1903-2003. J. Propulsion Power 2003, 19, 1089-1107.
4. Jackson, T. A.; Eklund, D. R.; Fink, A. J. High Speed Propulsion: Performance Advantage of Advanced Materials. J. Mater. Sci. 2004, 39, 5905-5913.
5. Edwards, T. Cracking and Deposition Behavior of Supercritical Hydrocarbon Aviation Fuels. Combust. Sci. and Tech. 2006, 178, 307-334.
6. Vajda, S.; Pellin, M. J.; Greeley, J. P.; Marshall, C. L.; Curtiss, L. A.; Ballentine, G. A.; Elam, J. W.;Catillon-Mucherie, S.; Redfern, P. C.; Mehmood, F.; Zapol, P. Subnanometre Platinum Clusters as Highly Active and Selective Catalysts for the Oxidative Dehydrogenation of Propane. Nat. Mater. 2009, 8, 213-216.
7. Goddard, S. A.; Cortright, R. D.; Dumesic, J. A. Deuterium Tracing Studies and Microkinetic Analysis of Ethylene Hydrogenation over Platinum. J. Catal. 1992, 137, 186-198.
8. Anderson, S. L.; Alexandrova, A. N.; Dumesic, J. A.; Mavrikakis, M.; Khanna, S. 5
N.; Winans, R. E.; Zare, R. N. Catalyst and Fuel Interactions to Optimize Endothermic Cooling; Air Force Office of Scientific Research, U.S. Government Printing Office: Washington, DC, 2016; 68
9. Tan, K. F.; Chang, J.; Borgna, A.; Saeys, M. Effect of Boron Promotion on the Stability of Cobalt Fischer–Tropsch Catalysts. J. Catal. 2011, 280, 50-59.
10. Xu, J.; Chen, L.; Tan, K. F.; Borgna, A.; Saeys, M. Effect of Boron on the Stability of Ni Catalysts During Steam Methane Reforming. J. Catal. 2009, 261, 158-165.
11. Heiz, U.; Schneider, W.-D. Nanoassembled Model Catalysts. J. Phys. D: Appl. Phys. 2000, 33, R85.
12. Goodman, D. W. Catalysis by Metals: From Extended Single Crystals to Small Clusters. Surf. Rev. Lett. 1994, 1, 449 -55.
13. Rainer, D. R.; Xu, C.; Goodman, D. W. Characterization and Catalysis Studies of Small Metal Particles on Planar Model Oxide Supports. J. Mol. Catal. A: Chem. 1997, 119, 307-325.
14. Henry, C. R. Surface Studies of Supported Model Catalysts. Surf. Sci. Rep. 1998, 31, 231235-233325.
15. Campbell, C. T.; Grant, A. W.; Starr, D. E.; Parker, S. C.; Bondzie, V. A. Model Oxide-Supported Metal Catalysts: Energetics, Particle Thicknesses, Chemisorption and Catalytic Properties. Top. Catal. 2000, 14, 43-51.
16. Lim, D.-C.; Hwang, C.-C.; Ganteför, G.; Kim, Y. D. Model Catalysts of Supported Au Nanoparticles and Mass-Selected Clusters. Phys. Chem. Chem. Phys. 2010, 12, 15172-15180
17. Habibpour, V.; Wang, Z. W.; Palmer, R. E.; Heiz, U. Size-Selected Metal Clusters: New Models for Catalysis with Atomic Precision. J. Appl. Sci. 2011, 11, 1164-1170.
18. Gao, F.; Goodman, D. W. Model Catalysts: Simulating the Complexities of Heterogeneous Catalysts. Annu. Rev. Phys. Chem. 2012, 63, 265-286.
19. Heiz, U.; Sanchez, A.; Abbet, S.; Schneider, W.-D. Tuning the Oxidation of Carbon Monoxide Using Nanoassembled Model Catalysts. Chem. Phys. 2000, 262, 189-200.
CHAPTER 2
ETHYLENE DEHYDROGENATION ON Pt4, 7, 8 CLUSTERS ON Al2O3: STRONG
CLUSTER SIZE DEPENDENCE LINKED TO PREFERRED CATALYST
MORPHOLOGIES
Reprinted with permission from Eric T. Baxter, Mai-Anh Ha, Ashley C. Cass, Anastassia
N. Alexandrova, Scott L. Anderson, and ACS Catalysis 7, 3322. Copyright 2017,
American Chemical Society 7
2.1 Overview
Catalytic dehydrogenation of ethylene on size-selected Ptn (n = 4, 7, 8) clusters deposited on the surface of Al2O3 was studied experimentally and theoretically. Clusters were mass-selected, deposited on the alumina support, and probed by a combination of low energy ion scattering, temperature-programmed desorption and reaction of C2D4 and
D2, X-ray photoelectron spectroscopy, density functional theory, and statistical mechanical theory. Pt7 is identified as the most catalytically active cluster, while Pt4 and
Pt8 exhibit comparable activities. The higher activity can be related to the cluster structure and particularly to the distribution of cluster morphologies accessible at the temperatures and coverage with ethylene in catalytic conditions. Specifically, while Pt7 and Pt8 on alumina have very similar prismatic global minimum geometries, Pt7 at higher temperatures also has access to single-layer isomers, which become more and more predominant in the cluster catalyst ensemble upon increasing ethylene coverage. Single- layer isomers feature greater charge transfer from the support and more binding sites that activate ethylene for dehydrogenation rather than hydrogenation or desorption. Size- dependent susceptibility to coking and deactivation was also investigated. Our results show that size-dependent catalytic activity of clusters is not a simple property of single cluster geometry but the average over a statistical ensemble at relevant conditions.
2.2 Introduction
The subnano clusters are known to have special catalytic reactivity.1-4 For example, a specific range of cluster size often results in preferential reaction pathways or significantly altered reactivity because in the subnano regime clusters are affected by 8
size-dependent electronic and geometric structure. One issue complicating this situation is that there may be multiple thermally accessible cluster structures with significantly different electronic and binding site properties. Here, we use a combination of experimental probes of cluster structure and binding site distributions together with density functional theory (DFT) to explore the range of thermally accessible structures and study the effects of size-dependent cluster structure on ethylene binding and dehydrogenation.
Platinum’s ability to (de)hydrogenate hydrocarbons is well-documented, and it is of interest to see if small clusters of Pt can be good dehydrogenation catalysts with useful selectivity that might enable more economical use of precious metals in catalysis. An important part of the problem is the stability of small clusters with respect to both sintering or agglomeration and deactivation by carbon deposition (coking).5-8
This work focuses on catalytic dehydrogenation on Pt4, Pt7, and Pt8 clusters supported on alumina. Significant differences have been found experimentally between
Pt7 and Pt8, here and elsewhere, and much of our effort is focused on understanding why.
Pt4 was included as an example of a smaller cluster. In this size range, strong size effects on activity have been noted. Vajda et al. observed Pt8–10’s activity of 40–100 times for oxidative dehydrogenation of propane;2 Roberts et al. observed strongly size-dependent activity for CO oxidation, which was correlated to changes in both the valence electronic structure and the number of CO binding sites on top of the clusters.9
There is evidence suggesting that such effects may be at least partly related to a structural transition occurring around Pt7 and Pt8. Low energy ion scattering (ISS) for
Ptn/alumina/Re(0001) showed an abrupt ∼15% drop in the Pt ISS signal going from Pt7 to
9
Pt8, indicating a transition to morphologies where fewer Pt atoms are in the ISS- accessible top layer of the clusters.9 As discussed below, we verified that the effect also occurs for the Ptn/alumina/Ta(110) system studied here. Indeed, the drop between Pt7 and
Pt8 is actually somewhat larger (∼24%) for alumina/Ta(110). The inference of a drop in the fraction of Pt in the surface layer at Pt8 is consistent with a scanning tunneling
10, 11 microscopy study of Ptn/TiO2 by Watanabe et al. who observed a transition from single- to multilayer clusters between Pt7 and Pt8. Such transitions change the number and type of adsorbate binding sites exposed on the clusters; however, it is important to recognize that adsorbate binding can drive cluster isomerization, i.e., it is necessary to characterize both adsorbate-free and adsorbate-covered structures. As shown below, Pt7 is able to adsorb more ethylene than Pt4 or Pt8, on either a per Pt atom or per cluster basis, consistent with additional Pt atoms exposed in the surface layer. DFT allows this effect and the earlier ISS observations to be explained.
2.3 Results and Discussion
We begin by discussing DFT results for adsorbate-free Pt7 and Pt8 on α-alumina, as summarized in Figure 2.1 and Table A.4. The charges on the Pt atoms in each structure are indicated and discussed below.
2.3.1 Cluster Catalyst Structures
Small Pt clusters have many structural isomers with similar energies,12 and both
Pt7/alumina and Pt8/alumina are found to have 5–7 isomers of very different geometries, predicted to be populated in the 450–700 K temperature range where dehydrogenation is
10 observed. At the elevated temperatures relevant to catalysis and in the limit of no kinetic trapping, strong structural fluxionality and the presence of several isomers are expected.13
Because catalytic properties may be dominated by any one or few of the isomers, it is important to consider all thermally relevant structures.14 We will generally describe the structures as being either single-layer, where all Pt atoms are exposed in the surface layer, or prismatic, where one or more atoms are buried under the cluster surface. The lowest energy isomers of Pt7 and Pt8 on alumina are shown in Figure 2.1 together with their
Bader charges and Boltzmann populations at 700 K. Pt7 is found to have both single layer and prismatic structures in the thermally accessible set, with prismatic geometries (global minimum and fourth isomer) comprising 66.7% of the Boltzmann population at 700 K and with the balance being single-layer geometries (second, third, and fifth isomers). In contrast, all of the accessible isomers of Pt8 are prismatic.
The calculated structures provide an explanation for the ISS observation that the
9 fraction of Pt in surface layer drops between Pt7 and Pt8. ISS was done at 130 K; thus, if the isomer distribution is equilibrated, only the lowest energy structures of each cluster would have significant populations. For Pt7, this structure is prismatic but exposes 6 of its
7 atoms (i.e., ∼86%) in the ISS-accessible surface layer. The lowest energy isomer of Pt8 exposes only 6 of its 8 atoms (75%) in the surface layer; thus, we would predict a
∼12.5% drop in ISS intensity between Pt7 and Pt8. It is not unlikely; however, that there are barriers to isomerization, such that some higher energy structures remain as the sample is cooled. To the extent that this kinetic trapping occurs, it would tend to give an even larger intensity drop between Pt7 and Pt8 because Pt7 has several single layer isomers where 100% of the Pt would be detectable, whereas all of the Pt8 isomers are 11
prismatic.
At higher temperatures, relevant to the temperature-programmed desorption and reaction (TPD/R) experiments, the diversity of geometries of Pt7 offers a richer set of binding sites for ethylene as opposed to the more uniform structures populated for Pt8, and this should be reflected in the chemical activity.
2.3.2 Size-Dependent Catalytic Activity
The chemical properties of the clusters, as probed by TPD/R, are summarized in
Figure 2.2 for two consecutive TPD/R experiments on samples containing Pt4, Pt7, and
Pt8. For each experiment, the samples were dosed with 5 L of C2D4 at 150 K and then cooled to ∼130 K prior to each TPD/R heat ramp (3 K/sec). Figure 2.2 shows results for the two species, C2D4 and D2, observed to have significant desorption signals. Desorption is reported in terms of C2D4 or D2 molecules desorbing per Pt atom per second, taking
14 2 advantage of the fact that we know the Pt loading quite precisely (1.5 × 10 /cm ). The D2 signals were corrected for the contribution from mass spectrometer cracking of desorbing
C2D4, and the uncorrected data are reported in Figure A.1. To avoid interference from high background signals at masses 2 and 28, most experiments were done with C2D4.
Experiments with C2H4 were also done to look for acetylene desorption; however, none was observed. In addition, no signal for ethane was observed, indicating that hydrogenation is negligible under these conditions.
C2D4 desorption from the cluster-free alumina/Ta(110) support is shown for comparison to the Ptn/alumina results. Only the result of the first TPD/R run is shown because the second run was identical. To allow direct comparison, the data for cluster-
12
free alumina were scaled as if these samples also contained the same amount of Pt as the
Ptn/alumina samples. For the cluster-free alumina film, C2D4 began to desorb at the TPD starting temperature with desorption peaking at ∼165 K and rapidly declining at higher temperatures. After correcting for the contribution from C2D4 cracking in the mass spectrometer, the D2 signal is zero, i.e., all ethylene adsorbed on the alumina film desorbs intact. The integrated number of C2D4 molecules desorbing from alumina is ∼7 ×
1012/cm2, i.e., on the order of 0.01 ML. The low intensity indicates that ethylene does not bind stably to most sites on the alumina film at 150 K but that there are a few stable binding sites, presumably corresponding to defects in the alumina surface. Even these defect sites bind C2D4 weakly such that it desorbs intact, well below room temperature.
For the cluster-containing samples, ethylene desorption also begins as the heat ramp is
+ started with a sharp peak near ∼165 K for all three cluster sizes. There is weak D2 signal at low temperatures (Figure A.1), but this is entirely due to dissociative ionization of desorbing C2D4. Both the temperature dependence and intensity of the 165 K peak match those for cluster-free alumina, indicating that this low temperature feature is simply due to C2D4 desorbing from defect sites on the alumina film. The fact that this low temperature component is not significantly affected by deposition of 0.1 ML equivalent of Ptn suggests that the clusters do not diffuse to and occupy these defect sites, at least in the <300 K range where Ptn deposition, C2D4 exposure, and desorption of the low temperature feature occur.
For Ptn-containing samples, there is also a broad C2D4 desorption component extending between ∼200 and 420 K, which clearly results from ethylene bound to the Pt clusters. The peak temperature of this component is ∼280 K for Pt4 and Pt8 and ∼300 K
13
for Pt7, and the intensity is also significantly higher for the Pt7-containing sample. Below
∼250 K (Pt4, Pt8) or ∼275 K (Pt7), only intact C2D4 desorption is observed, but at higher temperatures, D2 desorbs in a broad component extending to 650 K. In each case, the onset of D2 desorption is just below the peak C2D4 desorption temperature, as would be expected if dehydrogenation to generate D2 is in competition with C2D4 desorption.
Figure 2.1 also lists the integrated amounts of C2D4 and D2 observed to desorb from each sample, given in terms of number of molecules desorbing per Pt atom. Desorption/cm2 can be obtained simply by multiplying by the Pt coverage (1.5 × 1014 Pt atoms/cm2).
These numbers include desorption from both alumina and Ptn sites, and to compare the
Ptn-only desorption, it is necessary to subtract the alumina contribution, equivalent to
∼0.08 C2D4/Pt atom. Thus, in the first TPD/R experiment, the corrected desorption is
0.24 C2D4 and 0.13 D2 molecules per Pt atom for Pt4, compared to 0.30 C2D4 and 0.20 D2 for Pt7, and 0.23 C2D4 and 0.15 D2 for Pt8.
No additional D2 desorption was observed in select experiments where the temperature was ramped to 900 K, thus it is reasonable to assume that dehydrogenation is complete by 650 K. In that case, the total number of initially adsorbed C2D4 molecules per Ptn cluster can be estimated as the sum of the C2D4 desorption plus half the D2 desorption. This amounts to ∼0.3 C2D4/Pt atom for both Pt4 and Pt8, compared to 0.4
C2D4/Pt atom for Pt7. The numbers of C2D4 molecules initially adsorbed per cluster are
∼1.1, ∼2.7 and ∼2.4, respectively, for Pt4, Pt7, and Pt8. Thus, Pt7 provides significantly more binding sites than the other two cluster sizes on either a per atom or per cluster basis.
Study of small supported clusters is complicated by substrate-mediated 14 adsorption, in which molecules initially land on the alumina support, where they bind too weakly to be stable, diffuse, and bind stably to the Ptn. For our experiments with 0.1 ML equivalent Pt coverage, the effect is to substantially amplify the effective adsorbate exposure to the clusters, as will be demonstrated for C2D4 below. During the ∼20 min elapsing between the start of cluster deposition and the first TPD/R heat ramp, the clusters, on average, are exposed to ∼0.04 L of background CO, corresponding to ∼0.01
CO impacting per surface atom. During the first TPD/R run, CO desorption amounting to
∼0.5 CO molecules/cluster is observed, independent of cluster size. CO binds strongly to
Ptn (see below) and likely competes with C2D4 for Pt binding sites. Therefore, we expect that the integrated C2D4 numbers are somewhat lower than they would be if no CO were present.
As discussed above, a larger fraction of the Pt atoms is exposed in the surface layer of Pt7/alumina compared to that in Pt8/alumina, consistent with the observation that more C2D4 adsorbs on Pt7 than on Pt8. Clearly; however, understanding the TPD/R experiments requires consideration of how the Ptn isomer distribution is affected by ethylene adsorption and also of the factors that control branching between ethylene desorption and dehydrogenation.
Before discussing DFT results for ethylene-Ptn interactions, we consider the question of whether the temperature dependence observed for D2 desorption is controlled by the energetics of C2D4 decomposition or simply reflects the activation energy for desorption of D2. This point was tested by studying D2 TPD, and Figure A.2 compares the D2 desorption from separate samples of Pt8/alumina/Ta(110) after 5 L exposure to either D2 or C2D4 at 150 K. It can be seen that for the D2 exposure, desorption starts at 15
∼160 K compared to ∼220 K for the C2D4 exposure and is 90% complete by ∼400 K, at which point only about half the D2 from C2D4 has desorbed. In this temperature range, D2 desorption from Pt almost certainly involves recombination of absorbed D atoms; thus, the higher temperatures required to drive D2 desorption after C2D4 exposure suggest that the limiting factor is the activation energy for some step(s) in the C2D4 decomposition process rather than the D2 recombinative desorption energetics.
The desorption spectra were simulated to extract activation energies as described in Appendix A, which reports the best-fit energy distributions in Figure A.3. Simulation was based on assuming first-order kinetics for the limiting step, as might be expected for intact C2D4 desorption. On the basis of this assumption, the desorption energy for C2D4 bound on the alumina film is in the ∼0.5 eV range, while for C2D4 at Pt cluster sites, the desorption energy would range from ∼0.6 to 1.4 eV. For D2 production, under the assumption of a first-order limiting step, the activation energy would be in the 0.7–2.3 eV range. A combination of DFT and coverage-dependent TPD/R studies was used to probe
C2D4 adsorption and desorption, resulting in a more complex picture of the process.
Because theory on supported cluster systems is computationally demanding due to the large number of isomers and adsorption geometries involved, we focus our DFT work on ethylene adsorption and activation for dehydrogenation as the key processes influencing the kinetics for ethylene and hydrogen desorption. In addition, we chose the
Pt7/alumina system for the most in-depth work, both experimentally and theoretically.
Because the TPD results indicate that roughly three C2D4 molecules adsorb initially per
Pt7, we consider theoretically adsorption of one, two, and three ethylene molecules on the most important Pt7 isomers and also the factors that influence desorption vs
16
dehydrogenation.
The literature for ethylene binding and hydrogenation/dehydrogenation on various
Pt surfaces provides an important insight that aids interpretation of the DFT results.
Ethylene adsorption and decomposition has been extensively studied on various platinum surfaces using techniques such as TPD, reflection/absorption infrared spectroscopy
(RAIRS), and high-resolution electron energy loss spectroscopy (HREELS). At temperatures below 100 K, adsorbed ethylene forms di-σ bonds on close-packed
Pt(111)15, 16 and Pt(100)17 surfaces and π-bonds on the stepped sites of Pt(210) and (1 ×
1)Pt(110).18 On the close-packed surfaces, some of the di-σ bound ethylene desorbs intact at temperatures around 285 K; however, TPD of C2D4 and C2H4 coadsorbed on Pt(111) also yielded C2D3H and C2H3D, indicating that recombinative desorption of dissociatively chemisorbed ethylene also contributes to the ethylene desorption signal.19
At temperatures just above the ethylene desorption peak, H2 desorption begins, indicating the onset of dehydrogenation. The first dehydrogenation step results in formation of ethylidyne (≡CCH3), which has been shown to adsorb in 3-fold hollow sites by
HREELS20 and tensor LEED.21 Most studies consider ethylidene to be a spectator species.22-24 At higher temperatures, the ethylidyne undergoes further decomposition, giving rise to additional H2 desorption and going to completion by ∼700 K. For the stepped surface of Pt(210), some of the π-bound ethylene desorbs at ∼250 K, and then the remaining π-bound ethylene dehydrogenates at ∼300 K, resulting in desorption of H2 and formation of adsorbed ethylylidyne (≡CCH2–, both C atoms bound to the surface), which undergoes further decomposition, giving rise to additional H2 desorption and going to completion by ∼700 K. In contrast, upon heating the (1 × 1)Pt(110) surface to ∼160 K,
17
some of the π-bound ethylene is converted to di-σ bound ethylene. Between 270 and 330
K, the adsorbed ethylene reacts to form carbon atoms and ethylidyne on the surface accompanied by desorption of methane and H2. The remaining ethylidyne undergoes complete dehydrogenation by 450 K. Studies on alumina-supported Pt nanoparticles showed that at temperatures below 180 K, ethylene adsorbs in three distinct forms: π- bound ethylene, di-σ bound ethylene, and the ethylidyne species.25 By room temperature, all of the π-bound ethylene desorbs intact, while at higher temperatures, the remaining di-
σ bound ethylene is converted to ethylidyne.
From the perspective of interpreting the DFT results, the key insight from these studies is a correlation between the adsorbed configuration of ethylene and its subsequent reactivity.22, 24, 26-29 This correlation, which applies to both Pt surfaces and Pt clusters, is that π-bonded, sp2 configurations tend to result in hydrogenated products, while di-σ bonded, sp3 configurations result in dehydrogenated products. For our system, where no hydrogen is added and hydrogenation is not observed, we interpret this correlation as suggesting that the precursor to dehydrogenation is di-σ bonded ethylene, while π-bonded ethylene should tend to desorb intact. Our bonding analysis of ethylene adsorbed to Pt7 corresponds well to near-edge X-ray-absorption fine-structure (NEXAFS) studies on
Pt(111) with di-σ bound ethylene reflecting a bond-length of ∼1.5 Å and π-bound ∼1.4
Å.29 Moreover, bond angles of ≈120° and ≈97–115° reflect sp2 and sp3 hybridization present in adsorbed ethylene, respectively. We therefore will use the geometries calculated for adsorbed ethylene as indicators of the propensity to dehydrogenate.
Figure 2.3 shows the DFT results for ethylene binding to both the single layer and prismatic Pt7 isomers that were shown in Figure 2.1. The most stable structures (i) are
18 shown at the top, and additional local minima are shown below, with energetics and thermal populations summarized in Table 2.1. Recall that for adsorbate-free Pt7/alumina, the global minimum is prismatic; however, because the binding energy for the first ethylene molecule is ∼0.6 eV higher for the single-layer isomer, this becomes the global minimum for ethylene1/Pt7, and a variety of low energy ethylene1/Pt7 geometries based on the single layer isomer are shown in the left column. With one ethylene molecule adsorbed, the prismatic isomers shown in the second column are therefore local minima, stabilized by barriers associated with the considerable rearrangement required to convert to the single layer global minimum.
Bader charge analysis shows that ethylene adsorption is associated with electron transfer from Ptn to the carbon atoms of ethylene, suggesting that the charges on the adsorbate-free Ptn/alumina isomers should be related to their affinities for ethylene. The charges on each atom given in Figure 2.1 show that on average, Pt7 has greater electron
– transfer from the support compared to that of Pt8. Pt7 isomers take up 1.22–1.44 e
– compared to 0.80–1.24 e for Pt8. In addition, the single-layer Pt7 isomers have higher charge than the prismatic isomers and less uniform charge distributions, with some Pt atoms carrying the majority of the negative charge. Pt8, which has only prismatic isomers, has more uniform and lower charge distributions.
In principle, the extent of Ptn charging can be probed by XPS, and Pt 4d spectra for Pt7/alumina and Pt8/alumina are shown in Figure A.4. The stronger Pt 4f peaks were unusable because of Al 2p background. Although the 4d signal is noisy for 0.1 ML Pt coverage, it appears that the 4d binding energy (BE) for Pt7 is shifted ∼0.4 eV to higher energy compared to the Pt8 BE. We previously reported XPS BEs for Ptn on glassy 19
30 31 10 carbon and indium tin oxide, and Isomura et al. reported BEs for Ptn/TiO2(110). In all cases, the Pt7 BE is higher than that for Pt8, opposite to what might be expected if Pt7 is more negatively charged than Pt8. Note; however, that XPS BEs for small clusters are strongly affected by size-dependent final state effects,1, 32-34 and size-dependent rehybridization of metal orbitals has recently been identified as another factor in BEs for
35, 36 supported Pdn. As a result, interpreting the BE shift in terms of the initial state charge is not possible.
Upon ethylene binding, the calculated charge on the Pt binding site increases by
≈0.2–0.7 e, reflecting electron transfer from Pt to ethylene. Therefore, it is not unreasonable to expect that the more negatively charged cluster isomers and binding sites should tend to have higher ethylene binding energies. For example, the single-layer local minimum is 0.22 e– more negatively charged than the prismatic global minimum, and its ethylene adsorption energy is stronger by 0.68 eV. Therefore, both the larger average alumina-to-Ptn electron transfer for Pt7, and the existence of low-lying single-layer local minima that have the highest alumina-to-Ptn electron transfer, are consistent with the observation (Figure 2.1) that Pt7 binds ethylene more strongly than Pt8. The fact that Pt7 also binds more ethylene in saturation is also consistent with the larger fraction of Pt in the cluster surface layer. At low coverage, the prismatic global minimum Pt7 structure shows more di-σ ethylene binding (configurations ii and iv–vi, Figure 2.3, Table A.5) than the single-layer local minimum (only configuration iii). The di-σ bound ethylene often carries more negative total charge (ΔQethylene) as compared to its π-bound counterpart at all studied coverages (Table 2.1 and Tables A.5–A.7).
Because at the ethylene exposure temperature (150 K), the prismatic isomer of Pt7 20
should dominate, and because of the great computational expense, coverage-dependent ethylene binding was studied only for starting geometries based on this isomer. All possible adsorption sites (atomic, bridging, and hollow) were evaluated for this global minimum Pt7 isomer, and the six lowest energy geometries were used in further analysis.
The lowest energy minimum for one ethylene adsorbed on prismatic Pt7 was used as the starting geometry for adding the second ethylene, and the resulting two-ethylene minima were taken as starting geometries for adding the third. This is an approximation because the lowest energy geometry is not necessarily the precursor for higher coverage structures. Thus, there is some uncertainty as to whether the thermal populations in our coverage study include all important structures; however, the results provide at least qualitative insights.
With one or two ethylene molecules adsorbed, the prismatic Pt7 starting geometry is retained in the optimized structures, but for three adsorbed ethylene molecules, optimization from many of the prismatic starting geometries led to single layer isomers, dominating the thermally accessible ensemble (>78% Boltzmann populations at 450 and
700 K). As noted, even for the first adsorbed ethylene, the single layer structure is lower in energy than the prismatic isomer, but the prismatic isomer is stabilized by a barrier of unknown height. The same is likely true for two ethylene molecules, but clearly, the isomerization barrier vanishes when the third ethylene binds. Therefore, one inference from DFT is that adsorption of ethylene will tend to drive transition toward single layer isomers and that such structures are therefore likely to be more important in the experiments than would be suggested by consideration of only adsorbate-free Ptn geometries. Additionally, we observe that as ethylene coverage increases and cluster
21 geometries flatten, the populations of di-σ ethylene binding geometries, which are precursors to dehydrogenation, also increase from 7.35 to 84.26% at 450 K, i.e., at the peak of D2 desorption (Table 2.1).
A coverage-dependent TPD/R experiment was performed on Pt7/Al2O3 to gain further insight into the energetics and dynamics of ethylene desorption and the competition between desorption and dehydrogenation. Figure 2.4 compares C2D4 and D2 desorption from separately prepared Pt7/alumina samples exposed to 5, 0.1, and ∼0.01 L of C2D4 at 150 K, otherwise following the same procedure as in Figure 2.2 For comparison, C2D4 desorption from the cluster-free alumina film is also shown for 5 and
0.1 L C2D4 exposures. Figure 2.4 also gives the numbers of C2D4 and D2 molecules desorbing per Pt atom, calculated by subtracting the desorption from cluster-free alumina and then integrating.
For the Pt7/alumina samples, C2D4 desorption clearly is mostly from sites on alumina at the lowest temperatures and from Pt-associated sites at temperatures above
∼200 K. The Pt-associated desorption feature is quite broad, which normally would be taken as evidence for a wide distribution of desorption energies, as suggested by the fits to the TPD/R results discussed above. In that scenario, we would expect desorption to shift to higher temperatures for decreasing coverage, because in subsaturation coverages, adsorbates should tend to diffuse to and desorb from the most stable sites available.
Furthermore, if C2D4 in the strongest di-σ binding sites has the highest probability of decomposing rather than desorbing intact, we might expect that the branching to D2 should increase with decreasing coverage, as is observed from ∼32% in the 5 L exposure to ∼52% for the 0.01 L exposure. 22
The coverage dependence of the desorption temperatures do not fit this simple scenario, however. The C2D4 desorption spectrum is weakly coverage dependent, and if anything, there is less desorption at the highest temperatures for the lowest coverage.
Furthermore, while the upper and lower temperature limits for D2 desorption are independent of initial coverage, the peak of D2 production shifts to substantially lower temperatures for lower initial C2D4 coverage. Given the DFT results showing that the relative stability of different Pt7 isomers is dependent on C2D4 coverage and that the isomer distribution evolves with temperature, we believe that the measured desorption temperature distributions reflect complex dynamics involving changes in cluster structure as part of the C2D4 desorption and decomposition mechanism.
Figure 2.4 also illustrates the importance of substrate-mediated adsorption for
15 2 highly dispersed clusters. Exposure of 5 L corresponds to 1.8 × 10 C2D4 collisions/cm or ∼1.2 collisions/surface atom. If adsorption at Pt sites occurred only in C2D4 collisions on Pt7, reducing the C2D4 exposure substantially should substantially reduce the C2D4 coverage on Pt7. Assuming that adsorbed C2D4 either desorbs intact or generates two D2 molecules, and subtracting the contribution from the alumina sites, the initial C2D4 coverage on Pt7 in the 5 L dose is ∼2.7 per Pt7 cluster. For a dose 50 times lower, the coverage is ∼2.1 C2D4/Pt7, and for exposure 500 times lower, the initial coverage is still
∼0.6 molecules/Pt7.
2.3.3 Routes of Deactivation
From the perspective of the catalytic properties of small Ptn/alumina, it is important to understand how the clusters are modified by heating, adsorption, desorption, 23 and dehydrogenation of ethylene. The DFT results suggest that isomerization is likely during the TPD/R cycle, and the second TPD/R runs on each sample (Figure 2.2) indicate that irreversible changes also occur. The amount of C2D4 desorbing at high temperatures decreased in the second run with an offsetting increase in desorption at low temperatures.
The total amount of C2D4 desorbing in the second run was ∼0.3 molecules per Pt atom for all three samples, which is essentially identical to the amount observed in the first runs for Pt4 and Pt8. For Pt7; however, ∼0.3 C2D4/Pt atom represents a ∼25% drop compared to the first run. As shown above, C2D4 adsorbed on the alumina support all desorbs intact at low temperatures, and the results in the first and second TPD/R runs are identical for cluster-free alumina. For Ptn/alumina, it is reasonable to assume that the alumina contribution to the C2D4 signal is also identical in the first and second runs, thus implying that C2D4 desorption from Pt sites shifted to lower temperatures in the second
TPD/R run, i.e., the ethylene-Pt desorption energies substantially decreased. The temperature dependence for D2 desorption did not differ dramatically between the first and second TPD/R runs; however, the integrated amount of D2 dropped by ∼60–70%.
This behavior is what would be expected if the activation energy for C2D4 dehydrogenation is unchanged in the second run so that more of the C2D4, which is bound more weakly in the second run, desorbs at temperatures below the onset for decomposition.
In the second TPD/R run, the dependence on deposited cluster size is much weaker than that in the first, where Pt7 stands out. This change could indicate that thermal or adsorbate-induced ripening or sintering generates a size distribution that no longer depends on the deposited size; however, there are other possibilities. DFT suggests that 24 the larger amount, stronger binding, and greater propensity toward dehydrogenation of
C2D4 on Pt7 compared to that on Pt8 (based on its structural and electronic characteristics) is related to the existence of a larger number of strong di-σ binding sites on Pt7. If carbon left on the surface by D2 desorption in the first TPD/R run tends to poison the strong di-σ binding sites, this would reduce both the average C2D4 binding energy in the second
TPD/R run, and the amount of D2 produced, in line with observations. The fact that more
D2 is produced per Pt atom in the first TPD/R run for Pt7 than either Pt4 or Pt8 implies additional carbon poisoning for Pt7 in the second run, tending to bring its binding/reactivity properties more in line with those of Pt4 and Pt8.
Several experiments were done to provide additional insight into how TPD/R changes the clusters. CO binds strongly to Pt and weakly to alumina, providing an alternative probe of the effect of different experimental manipulations on the availability of Pt binding sites. Figure 2.5 compares CO TPD for a set of Pt7/alumina/Ta(110) samples that were each first exposed to a particular manipulation and then probed by CO
TPD (10 L 13CO exposure at 150 K, heating at 3 K/sec to 700 K). Little CO desorbs from the alumina support, but for as-deposited Pt7/alumina, strong bimodal CO desorption is observed with a low temperature component peaking at ∼165 K and a broader high temperature component peaking at ∼520 K. Simply heating Pt7/alumina to 700 K in vacuum results in a ∼40% decrease in high temperature CO desorption and an increase in low temperature desorption. We previously studied CO TPD from Ptn/alumina/Re(0001)
9 (2 ≤ n ≤ 18) with similar results to the Pt7/alumina/Ta(110) TPD shown here. We found that the high temperature CO desorption intensity during the first TPD on
Ptn/alumina/Re(0001) increased substantially with increasing cluster size. Therefore, we 25
can conclude that the decrease in high temperature CO desorption following 700 K heating in Figure 2.5 cannot be explained by thermal sintering or ripening alone. We are not claiming that sintering/ripening are unimportant, but there must be other changes as well. For example, 700 K annealing may cause changes in the as-deposited isomer distribution.
After a single C2D4 TPD/R run, there is a 55% decrease in the high temperature
CO desorption, and the decrease is ∼75% for CO TPD following 6 C2D4 TPD/R runs.
Both heating and C2D4 TPD/R result in an increase in low temperature CO desorption, but overall, the total amount of CO desorbing decreased by ∼25% after 700 K heating and ∼40 and 50%, respectively, for 1 and 6 ethylene TPD/R runs. The larger effect of
C2D4 TPD/R compared to that of 700 K heating is attributed to carbon left on the surface by D2 desorption, blocking the Pt binding sites associated with high temperature CO desorption.
To provide additional insight into how ethylene binds to Ptn/alumina and the effects of heating and carbon deposition, we did two types of He+ ISS experiments.
Figure 2.6 compares raw ISS data for Pt7/alumina after a variety of experimental operations. The peaks primarily result from scattering of He+ from single Pt, O, and Al atoms in the top layer of the sample. He+ signal from multiple or subsurface scattering
37, events is strongly attenuated, contributing mostly to the weak background at E/E0 ≤ 0.6.
38 Small Pt clusters, where most or all of the Pt atoms are in the surface layer, should give large Pt ISS signals and also cause some attenuation of ISS signals from the alumina support, although the attenuation should be small because the Ptn coverage is low.
Isomerization or agglomeration of clusters to form multilayer structures reduces the
26
fraction of Pt in the surface layer, which should appear as a drop in Pt ISS signal.
Similarly, adsorbates binding on top of the clusters attenuate the Pt ISS signal, while adsorbates binding on the alumina or around the cluster periphery have little effect on Pt signal but may attenuate signal from alumina.
In the top frame of Figure 2.6, ISS data are compared for the cluster-free alumina film, as-deposited Pt7/alumina, and Pt7/alumina that was heated to 700 K in ultrahigh vacuum (UHV). Note the presence of a small peak at E/E0 ≈ 0.9 for cluster-free alumina, attributed to a ∼1% concentration of Ta in the surface layer from diffusion during high
39, 40 temperature alumina growth on the Ta(110) substrate. For samples with Ptn deposited, this Ta signal is presumably still present, underlying the much stronger Pt peak. Because the Ta intensity is so small, we have not attempted to subtract it.
When as-deposited Pt7/alumina is heated to 700 K, there is a small increase in Pt
ISS intensity. As discussed above, TPD shows that the as-deposited clusters are decorated with ∼0.5 adventitious CO molecules per cluster, and Figure A.6 implies that these CO molecules bind such that they attenuate ISS signal from the Pt clusters. Using the extrapolation procedure illustrated in Figure A.5 and detailed previously,9, 41, 42 we estimate the attenuation to be ∼30%, and the star in Figure 2.6 indicates the estimated value for adsorbate-free Pt7/alumina. CO desorbs by 700 K (Figure 2.5), which should restore the Pt ISS intensity; thus, the fact that only a small signal increase occurs implies that heating also drives morphology changes that offset the expected increase. From the size of the offset, we can rule out formation of large three-dimensional particles, but thermal isomerization from single layer to prismatic isomers or modest ripening of the cluster size distribution are possible.
27
The lower frame of Figure 2.6 compares the effects of C2D4 exposure and TPD/R.
One sample was exposed to 5 L of C2D4 at 150 K and then probed by ISS while cold, resulting in Pt ISS attenuation by ∼90% compared to the adsorbate-free limit,
+ demonstrating that C2D4 adsorbs in geometries that strongly attenuate He signal from Pt.
The Al and O ISS signals are attenuated by much smaller amounts, consistent with the
TPD data, indicating that little C2D4 adsorbs on alumina at this temperature. ISS data are also shown for a Pt7/alumina sample after a single C2D4 TPD/R run under the conditions of Figure 2.2 and for another sample run through 6 consecutive TPD/R runs prior to ISS analysis. After one TPD/R run, the Al and O intensities recover to the pre-exposure values, but the Pt ISS intensity remains ∼15% below the as-deposited value or ∼45% below the adsorbate-free limit. This post TPD/R value is ∼25% smaller than that measured after 700 K heating, and this additional attenuation is not surprising given that we know that carbon is left on the surface by D2 desorption (∼1.4 C atoms/Pt7). Figure
A.7 gives the integrated D2 desorption signal during 6 sequential C2D4 TPD/R runs, allowing us to estimate that a total of ∼3.4 C atoms are left behind per initially deposited
Pt7. If this carbon remains on top of Pt, it would cause at least a substantial fraction of the
∼65% Pt ISS attenuation observed after 6 TPD/R runs, but the attenuation may also reflects sintering or other changes in the Pt morphology. We also probed the residual carbon by XPS. No C 1s signal was detected after one or two TPD/R runs, but as shown in Figure A.8, after six runs, C 1s signal was observed, albeit too weak for accurate quantitation.
Temperature-dependent ISS (TD-ISS) provides more detailed information about the nature of the adsorbate binding on Pt7. TD-ISS is essentially a C2D4 thermal 28
desorption experiment in which a Pt7/alumina sample was dosed with 5 L of C2D4 at 150
K and then characterized by ISS. The sample temperature was then increased in 50 K steps, with ISS measurements at each step. Figure 2.7 plots the Pt ISS intensities, normalized to the sum of Al and O intensities, as a function of temperature (open circles).
The top axis gives the cumulative He+ exposure to the sample at the time the Pt ISS peak was being measured at each temperature. For comparison, the C2D4 and D2 desorption data from Figure 2.2 are superimposed, and a horizontal solid line indicates the expected intensity for adsorbate-free Pt7, estimated as shown in Figure A.6. Comparing the first point at 150 K to the value for adsorbate-free Pt7/alumina, we see that 5 L C2D4 exposure at 150 K resulted in attenuation by ∼93%, essentially the same attenuation seen in the raw ISS data in Figure 2.6. From TPD, we know that this initial exposure leads to adsorption of ∼2.7 C2D4 molecules associated with the clusters with some additional
C2D4 bound on the alumina. The expectation is that as the sample is heated and C2D4 desorbs or decomposes, the Pt ISS signal should recover.
To interpret the results quantitatively, it is necessary to understand how He+ sputtering of Pt and C2D4 affects the Pt ISS signal. The extrapolation experiment (Figure
A.5) also gives the decay rate of Pt ISS signal as a function of He+ exposure, and this is plotted in Figure 2.6 as a gray dashed line labeled “Pt signal loss from sputtering”. For
+ C2D4-covered Pt7/alumina, He sputter removal of C2D4 will tend to increase the Pt signal, and this rate was measured in an experiment where a sample was dosed with 5 L of C2D4 at 150 K and then repeatedly probed by ISS while held at 150 K (green dotted line labeled “Pt signal recovery by sputtering”). The Pt ISS intensity just after a single
TPD/R cycle (Figure 2.7) is indicated on the right axis by a red star. 29
As shown by the superimposed TPD/R data, by 200 K, the lowest temperature
C2D4 component has desorbed, but there is no recovery of Pt ISS signal beyond that expected from C2D4 sputtering. By 400 K, most of the C2D4 desorption and ∼30% of D2 desorption should have occurred, i.e., 85% of the initial C2D4 should have either desorbed or decomposed, but the Pt signal recovered to only ∼30% of the adsorbate-free value. By
550 K, ∼95% of the total amount of C2D4 and D2 desorption should have occurred, but the Pt ISS signal was still ∼35% below that expected for adsorbate-free Pt7/alumina. Note that the intensity at this point is essentially identical to that observed immediately after a
TPD/R experiment (red star). The Pt ISS intensity continued to increase and then leveled off above ∼600 K at a value well above that seen after a TPD/R run (red star) but ∼10% below the value that would be expected based on the “Pt signal loss from sputtering” trend line.
We interpret the results as follows. The TPD component below 200 K is associated with C2D4 bound on alumina (Figure 2.2); thus, its desorption is not expected to have any effect on the Pt ISS signal, as observed. By 300 K, ∼50% of the initial C2D4 has desorbed, including a significant fraction of the Pt-associated C2D4, but there is only modest recovery of Pt signal, indicating the weakly bound C2D4 is in sites where it does not strongly attenuate Pt ISS signal. Only as the more strongly bound C2D4 desorbs or decomposes at higher temperatures does the Pt signal recovery accelerate, indicating that this strongest C2D4 binding component is in sites that are efficient at attenuating ISS from
Pt7. TPD/R shows that it is this strongly bound C2D4 that is most likely to decompose, generating D2. Given our 45° angle of incidence and detection along the surface normal, we expect that these sites should be generally on top of the clusters. The fact that Pt ISS
30
recovery reaches only 90% of the adsorbate-free limit is not surprising because we know that carbon is left on the surface by D2 desorption. Indeed, the substantially lower Pt ISS signal measured after a TPD/R run (red star) suggests that without the effect of He+ sputtering throughout the TD-ISS run, even more decomposition products are left on the cluster surface.
During an earlier study of CO interactions with Ptn/alumina/Re(0001), we measured but did not publish a TD-ISS study of CO binding for Pt4/alumina/Re(0001), and this data is shown in Figure A.6 for comparison. The aspect that is relevant to the
C2D4 results here is that the experiment shows that for CO, the strongest binding sites are also those which cause the largest Pt ISS attenuation, i.e., sites on top of the Pt clusters. A
41 similar conclusion was reached in TD-ISS studies of CO on Pdn/TiO2 and Pdn/alumina.
Finally, DFT was also used to examine carbon atom binding to Pt7 and Pt8 to determine the most stable binding geometries and also to see if the propensity for coking has a role in the observed efficiency of dehydrogenation on these clusters. Shaikhutdinov et al. noted that in alumina-supported Pd catalysts, carbon deposits began to form at ca.
550 K from di-σ bound ethylene.24 As a first approach to understanding coking, we analyzed C-sticking energetics for isomers of deposited Pt clusters whose Boltzmann populations sum to >99% at 700 K. By summing the C-sticking energies for the
Boltzmann-weighed populations for Pt7 and Pt8 (i.e., ∑P EC), we obtain an estimate of the coking susceptibility of the isomer ensemble for each cluster size. Higher affinity to C should also correlate with a lower barrier to the dehydrogenation vs desorption.
Pure Pt clusters succumb to coking at higher temperatures due to the increasing population of isomers with very little resistance to carbon deposits (Figure 2.8). For Pt7,
31
∑P EC decreases with increasing temperature from −7.30 eV at 450 K to −7.38 eV at 700
K. For Pt8, ∑P EC remains high at > −8.0 eV (see Table 2 for details). This suggests that
Pt8 should undergo coking more readily. The electrophilic C pulls electrons from the Pt clusters, resulting in a ΔQC of −0.40 to −0.56 eV (Figures A14 and A15). C preferentially adsorbs on a hollow site with 3–4 Pt–C bonds and prefers the more electron-rich isomers of Pt7 (Figure 2.1, isomers II–IV) and Pt8 (Figure 2.1, isomers I–IV). We note that catalyst deactivation is a complicated process that may involve the buildup of the C-rich deposits, cluster ripening, and more dramatic restructuring, and it is not fully captured by theory. Experimentally, it is clear that Pt7 deactivates more strongly after the first TPD run.
2.4 Conclusions
We reported on ethylene dehydrogenation on size-selected, alumina-deposited subnano Pt clusters accessed via a combination of experiment and theory. Remarkably, deposited Pt7 is found to be significantly more active than deposited Pt8 and Pt4, which in turn have comparable activities. Pt7 also deactivates more easily through a number of potential ways, including coking and ripening. Throughout this study, we found that understanding many aspects of the experimental results requires consideration of the accessible ensemble of cluster isomers and how this evolves with C2D4 coverage and temperature. For example, the higher C2D4 binding affinity and dehydrogenation branching for Pt7 compared to those of Pt8 can be recovered only if multiple cluster minima are considered. Furthermore, the importance of single-layer geometries becomes obvious only after realistic coverage is included because C2D4 binding drives a transition
32 to single layer isomers where binding is stronger and more likely to result in dehydrogenation. From the dependence of ethylene activation by more negatively charged Pt atoms, it can be proposed that surfaces that charge Pt cluster more should be better supports for Pt catalysts for dehydrogenation. In addition, pronounced differential affinity for C binding is seen only in the ensemble. These results call for a change in paradigm when subnano cluster catalysts are characterized computationally, tracking isomer distributions with and without the adsorbate(s) of interest.
2.5 Methodology
2.5.1 Experimental Section
The experiments were performed using a cluster deposition/surface analysis instrument described previously,41, 43 which allows in situ sample preparation and characterization. Briefly, the instrument consists of a laser vaporization cluster ion source that feeds into a mass-selecting ion deposition beamline that terminates in a UHV chamber (base pressure ∼1.5 × 10–10 Torr). The main UHV chamber is equipped for sample cleaning and annealing and houses a differentially pumped mass spectrometer for
TPD/TPR studies and hardware for sample characterization by XPS and low energy ISS.
The model catalyst supports were prepared alumina films grown on a 7 × 7 mm
Ta(110) single crystal (Princeton Scientific Corporation), which was spot-welded to tantalum heating wires that were attached to a liquid-nitrogen cooled cryostat mounted at the end of a manipulator. The sample could be cooled to ∼120 K and resistively heated to
∼1200 K. A filament mounted directly behind the sample that allowed heating by electron bombardment to temperatures greater than 2100 K. Sample temperature was 33
monitored by a type C thermocouple spot-welded to the back side of the crystal. Because type C thermocouples have low output at temperatures below 300 K, the temperature scale was calibrated by temporarily attaching an additional type K thermocouple, with the result that the two thermocouples agreed to within 3 K over the 120–1000 K range where type K can be used.
Alumina thin films were grown using procedures adapted from work of the
Goodman44-46 and Madey47, 48 groups. Aluminum was evaporated from a crucible
–6 16 mounted normal to the Ta(110) surface in 5 × 10 Torr O2 background pressure while holding the sample temperature at 970 K. Film thicknesses were determined for each sample from the Al 2s and Ta 4d XPS intensities, and for these studies, the growth rate was maintained at ∼2 Å/min. As discussed by Chen and Goodman, thin (∼1.5 nm) alumina films grown on Ta(110) show slightly distorted hexagonal symmetry attributed to either the (0001) or (111) face of α-alumina.49 We studied the effects of alumina thickness on the core and valence electronic properties of alumina grown on both
39 40 Ta(110) and Re(0001) and on the CO oxidation activity of Pdn clusters supported on alumina/Ta and alumina/Re. Because we found that properties became thickness- independent only above ∼3 nm, we used 3–6 nm thick films in the present study. Note that all experiments were carried out on freshly prepared samples to avoid issues of sample contamination or damage.
Model catalyst preparation began by cooling the cryostat and sample holder until the surface temperature reached 130 K and then flashing it to ∼2100 K for 5 min to remove any contaminants (including the previous alumina film) and annealing the crystal.
XPS and ISS of the surface after this heat treatment showed no contamination with the
34
exception of submonolayer amounts of surface oxygen. The sample was then lowered into a small UHV-compatible antechamber, where it was isolated from the main chamber by a triple differentially pumped seal to the cryostat. The antechamber was then flooded
–6 with 5 × 10 Torr of O2, and the alumina film was grown.
Following XPS characterization of the alumina film, the sample was flashed from
∼120 to 800 K to desorb any adventitious adsorbates that might have adsorbed during
XPS. To minimize exposure of the deposited clusters to background gases, deposition of mass-selected Ptn (n = 4, 7, or 8) clusters was done as the sample cooled back to 120 K, beginning when the sample reached 300 K. During deposition, the sample was positioned directly behind a 2 mm in diameter exposure mask, which defined the size of the cluster spot on surface. The Ptn coverage was monitored via the neutralization current of the soft landed (∼1 eV/atom) clusters on the support, and deposition was terminated for all samples such that they all had identical Pt loading of 1.5 × 1014 atoms/cm2 (∼0.1 ML), differing only in the size of clusters deposited. Deposition took 5–15 min.
For TPD/R measurements, a differentially pumped mass spectrometer (UTI 100 C with Extrel electronics) was used, viewing the main chamber through the ∼2.5 mm diameter aperture in the tip of a skimmer cone. The skimmer cone is surrounded by four
6 mm diameter dosing tubes that point at the sample position and can be connected to either continuous or pulsed valves. For dosing, the sample was positioned with the cluster spot centered on the skimmer aperture with a 2 mm separation to allow line of site from the dosing tubes to the cluster area. Calibration experiments show that the gas exposure to the cluster spot is ten times greater than the exposure to the chamber walls. For ethylene TPD/R experiments, the samples were exposed to 5 L of C2D4 at 150 K sample
35
temperature, chosen to minimize adsorption on the alumina support. The sample was then moved to 0.5 mm distance from the skimmer aperture, cooled to 135 K, and then ramped to 700 K at 3 K/sec while monitoring masses of interest desorbing from the surface. To examine the effects of heating and adsorbate exposure on the clusters, the TPD/R experiment (with fresh ethylene exposure) was repeated multiple times on each sample.
Select experiments were done under identical conditions but with C2D4 exposures of 0.1 and 0.01 L.
13 Because CO binds strongly to Ptn but not to alumina, we also did CO TPD experiments to investigate the effects of heating and ethylene decomposition on the availability of Pt binding sites. These experiments were carried out by exposing samples to 10 L of 13C16O at 150 K and then ramping the temperature from 135 to 700 K at 3
K/sec while monitoring desorption of 13CO and other masses of interest. Because of substrate-mediated adsorption,38 highly dispersed Pt clusters are also efficient at collecting adventitious CO, present in the chamber background at ∼5 × 10–11 Torr. After correcting the mass 28 TPD signal for C2D4 cracking in the ion source, the amount of CO adsorbed onto the clusters was found to be ∼0.5 CO molecules per cluster for Pt4, Pt7, and Pt8, i.e., approximately half of the clusters have one CO molecule adsorbed, desorbing above 500 K. This adventitious signal is independent of whether the sample was dosed with C2D4, i.e., C2D4 is not able to displace CO from the clusters. The amount of adventitious CO desorbing from a Pt-free alumina film sample is negligible.
To convert the ion signals measured during TPD/R to absolute numbers of molecules desorbing from the surface, we calibrated the mass spectrometer sensitivity in several ways.9, 50 Several times during the course of the experiments, we checked the
36
calibration of C2D4 and other gases of interest by filling the main UHV chamber with those gases to a measured pressure (correcting for ionization gauge sensitivity) while measuring the resulting ion signals. This results in a well-known flux of molecules effusing through the 2.5 mm diameter skimmer cone aperture into the mass spectrometer ion source (creating a known number density), allowing us to calculate the calibration factor for each gas. To check for possible changes in electron multiplier gain, this calibration was done daily for argon gas. The accuracy of this calibration approach was checked against calibrations based on desorption of saturated CO layers of known coverage from Pd(111) or Ni(110).41 We estimate that the calibration should be accurate to ±30%, mostly because of uncertainties in the angular distributions for desorption from clusters and the ionization efficiency vs angle.
Low energy ISS was used to observe the effects of cluster size, adsorbate binding, and TPD/R on the fraction of Pt atoms in the surface layer. ISS was done by loosely focusing a beam of 1 keV He+ onto the sample at 45° angle of incidence with an energy of 1 keV onto the sample and measuring the energy of He+ scattered along the surface normal. Peaks in the resulting energy spectrum are due to scattering of He+ from single atoms in the sample surface layer, identifying the masses of those atoms. Multiple scattering or subsurface scattering events contribute to a broad background, which is weak due to low ion survival probability in such trajectories.37 Because ISS is not a nondestructive technique, it was either done on separately prepared samples or on samples at the end of experimental sequences. 37
2.5.2 Computational
Because the alumina film used in the experiments is structurally similar to α- alumina(0001),49 all calculations were done for this surface, and the calculations focused on Pt7 and Pt8 because these showed interesting differences in the experiments. Plane wave density functional theory calculations of both gas-phase and adsorbed Pt7 and Pt8 were performed using Vienna Ab initio Simulation Package (VASP)51-54 with projector augmented wave potentials55 and the PBE56 functional. Bulk calculations were performed with a 8 × 8 × 3 Monkhorst–Pack k-point grid with large kinetic energy cutoffs of 520.0 eV and a stringent SCF (geometric) convergence criteria of 10–6 (10–5) eV, resulting in an optimized lattice constant of a = 4.807 Å and c = 13.126 Å for α-Al2O3 (0001), a slight increase compared to experiment.57, 58 This overestimation is typical of GGA functionals and corresponds to <0.1 Å increase in lattice constants. The α-alumina slab was modeled as a 3 × 3 unit cell with a vacuum gap of 15 Å and the bottom half of the slab kept fixed.
For calculations presented in this paper, large kinetic energy cutoffs of 400.0 eV and convergence criteria of 10–5 (10–6) eV for geometric (electronic) relaxations were employed. Only the most thermodynamically stable, Al-terminated surface was explored with an inward relaxation of 89.7% of the surface Al and O layers. Reproducing experimental results of −51 to −63% relaxation would require hydroxylation of the surface and introduce even more permutations of adsorbed cluster configurations.59, 60
Thus, this is beyond the scope of the current study.
Adsorbed structures were formed from the deposition of the lowest 5–6 gas phase structures under PBE levels of theory per manum with a thorough sampling of cluster faces to possible binding sites. Gas phase structures of Pt8 were found with the Adaptive
38
Force Field Coalescence Kick (AFFCK),61 an adaptive global minimum and local minima
62 search based on the Coalescence Kick (CK). For Pt7, structures from a study by Tian et al. were further optimized under VASP/PBE levels of theory, resulting in a new structure
(isomer II in Figure A.9 and Tables A.1 and A.2).63 A CK search also uncovered a new configuration, isomer III (see Figure A.9 and Table A.2). Note that the order of clusters composed of seven or more atoms will often be DFT method-dependent.61 This is further discussed in detail in the Supporting Information, utilizing the TURBOMOLE V6.6 program with def2-TZVP basis and both pure (hybrid) versions of the functionals, PBE
(PBE0) and TPSS (TPSSh), respectively.
The relevant equations regarding formation (Eform), adsorption (Eads), and reagent
(Ereag) energies may be found in Appendix A and follow the conventions presented in previous studies.64, 65 Appendix A also includes the relevant equations utilized for statistical and bonding analysis such as the Boltzmann probability for ith configuration
(Pi) and the Gibbs’ entropy (SG). SG allows us to estimate at a specific temperature T the entropic contribution (TSG) to the Helmholtz free energy (F = U – TSG). To evaluate the ensemble effects of local minima at Ptn, the summation of the Boltzmann-weighted adsorption energies (∑PTEads = ∑iPi,TEi,ads) and carbon-sticking energies at a temperature
T were calculated (∑PTEc = ∑iPi,TEi,c). For the coverage study of ethylene, the calculated adsorption of ethylene took on the forms:
E1 ethylene = E1 ethylene+Pt7ads − E1 ethylene,gas − EPt7ads
E2 ethylene = E2 ethylene+glob,Pt7ads − E1 ethylene+glob,Pt7ads − E1 ethylene,gas
E3 ethylene = E3 ethylene+glob,Pt7ads − E2 ethylene+glob,Pt7ads − E1 ethylene,gas.
Details of computational methods, isomers of gas phase and deposited clusters,
39 clusters with 1-3 adsorbed ethylene molecules, and C with charges, populations, energies, and other properties are given in Appendix A, Figures A9-A15, Tables A.1-A.7.
2.6 Acknowledgments
This work was supported by the Air Force Office of Scientific Research under a
Basic Research Initiative grant (AFOSR FA9550-16-1-0141) to A.N.A. and S.L.A. M.-
A.H. acknowledges the UCLA Department of Chemistry and Biochemistry Dissertation
Year Fellowship. CPU resources at the DoD High Performance Computing
Modernization Program (the United States Air Force Research Laboratory DoD
Supercomputing Resource Center (AFRL DSRC), the United States Army Engineer
Research and Development Center (ERDC), and the Navy DoD Supercomputing
Resource Center (Navy DSRC)) supported by the Department of Defense, XSEDE, and the UCLA-IDRE cluster were used to conduct this work.
2.7 References
1. Kaden, W. E.; Wu, T.; Kunkel, W. A.; Anderson, S. L. Science 2009, 326, 826– 829
2. Vajda, S.; Pellin, M. J.; Greeley, J. P.; Marshall, C. L.; Curtiss, L. A.; Ballentine, G. A.; Elam, J. W.; Catillon-Mucherie, S.; Redfern, P. C.; Mehmood, F.; Zapol, P. Nat. Mater. 2009, 8, 213– 216
3. Lee, S.; Molina, L. M.; Lopez, M. J.; Alonso, J. A.; Hammer, B.; Lee, B.; Seifert, S.; Winans, R. E.; Elam, J. W.; Pellin, M. J.; Vajda, S. Angew. Chem., Int. Ed. 2009, 48, 1467– 1471
4. Kwon, G.; Ferguson, G. A.; Heard, C. J.; Tyo, E. C.; Yin, C.; DeBartolo, J.; Seifert, S.; Winans, R. E.; Kropf, A. J.; Greeley, J.; Johnston, R. L.; Curtiss, L. A.; Pellin, M. J.; Vajda, S. ACS Nano 2013, 7, 5808– 5817
5. Campbell, C. T.; Campbell, J.; Dalton, P.; Henn, F.; Rodriguez, J.; Seimanides, S. 40
J. Phys. Chem. 1989, 93, 806– 814
6. Pettiette-Hall, C. L.; Land, D. P.; McIver, R. T., Jr; Hemminger, J. C. J. Am. Chem. Soc. 1991, 113, 2755– 2756
7. Hansen, T. W.; DeLaRiva, A. T.; Challa, S. R.; Datye, A. K. Acc. Chem. Res. 2013, 46, 1720– 1730
8. Brizuela, G.; Hoffmann, R. J. Phys. Chem. A 1998, 102, 9618– 9624
9. Roberts, F. S.; Kane, M. D.; Baxter, E. T.; Anderson, S. L. Phys. Chem. Chem. Phys. 2014, 16, 26443– 26457
10. Isomura, N.; Wu, X.; Hirata, H.; Watanabe, Y. J. Vac. Sci. Technol., A 2010, 28, 1141– 1144
11. Watanabe, Y.; Wu, X.; Hirata, H.; Isomura, N. Catal. Sci. Technol. 2011, 1, 1490– 1495
12. Zhai, H.; Alexandrova, A. N. J. Chem. Theory Comput. 2016, 12, 6213– 6226
13. Zhai, H.; Alexandrova, A. N. ACS Catal. 2017, 7, 1905– 1911
14. Gao, M.; Lyalin, A.; Takagi, M.; Maeda, S.; Taketsugu, T. J. Phys. Chem. C 2015, 119, 11120– 11130
15. Steininger, H.; Ibach, H.; Lehwald, S. Surf. Sci. 1982, 117, 685– 698
16. Ibach, H.; Lehwald, S. J. Vac. Sci. Technol. 1978, 15, 407– 415
17. Hatzikos, G.; Masel, R. Surf. Sci. 1987, 185, 479– 494
18. Yagasaki, E.; Backman, A. L.; Masel, R. I. Vacuum 1990, 41, 57– 59
19. Janssens, T. V. W.; Zaera, F. Surf. Sci. 1995, 344, 77– 84
20. Kesmodel, L.; Dubois, L.; Somorjai, G. Chem. Phys. Lett. 1978, 56, 267– 271
21. Starke, U.; Barbieri, A.; Materer, N.; Van Hove, M.; Somorjai, G. Surf. Sci. 1993, 286, 1– 14
22. Neurock, M.; van Santen, R. A. J. Phys. Chem. B 2000, 104, 11127– 11145
23. Anderson, A. B.; Choe, S. J. Phys. Chem. 1989, 93, 6145– 6149
24. Shaikhutdinov, S. K.; Frank, M.; Bäumer, M.; Jackson, S. D.; Oldman, R. J.; 41
Hemminger, J. C.; Freund, H.-J. Catal. Lett. 2002, 80, 115– 122
25. Mohsin, S. B.; Trenary, M.; Robota, H. J. J. Phys. Chem. 1988, 92, 5229– 5233
26. Perry, D. A.; Hemminger, J. C. J. Am. Chem. Soc. 2000, 122, 8079– 8080
27. Carlsson, A.; Madix, R. J. Chem. Phys. 2001, 115, 8074– 8082
28. Paz-Borbón, L. O.; Hellman, A.; Thomas, J. M.; Grönbeck, H. Phys. Chem. Chem. Phys. 2013, 15, 9694– 9700
29. Stöhr, J.; Sette, F.; Johnson, A. L. Phys. Rev. Lett. 1984, 53, 1684
30. Proch, S.; Wirth, M.; White, H. S.; Anderson, S. L. J. Am. Chem. Soc. 2013, 135, 3073– 3086
31. von Weber, A.; Baxter, E. T.; Proch, S.; Kane, M. D.; Rosenfelder, M.; White, H. S.; Anderson, S. L. Phys. Chem. Chem. Phys. 2015, 17, 17601– 17610
32. Chusuei, C. C.; Lai, X.; Luo, K.; Goodman, D. W. Top. Catal. 2000, 14, 71– 83
33. Bagus, P. Chemical information from XPS binding energy shifts: A unified view. Abstracts of Papers of the American Chemical Society; American Chemical Society: Washington, DC, March 23–27, 2003; pp U676– U677.
34. Dai, Y.; Gorey, T. J.; Anderson, S. L.; Lee, S.; Lee, S.; Seifert, S.; Winans, R. E. J. Phys. Chem. C 2017, 121, 361– 374
35. Kaden, W. E.; Büchner, C.; Lichtenstein, L.; Stuckenholz, S.; Ringleb, F.; Heyde, M.; Sterrer, M.; Freund, H.-J.; Giordano, L.; Pacchioni, G.; Nelin, C. J.; Bagus, P. S. Phys. Rev. B: Condens. Matter Mater. Phys. 2014, 89 (115436) 115431– 115438
36. Roberts, F. S.; Anderson, S. L.; Reber, A. C.; Khanna, S. N. J. Phys. Chem. C 2015, 119, 6033– 6046
37. Rabalais, J. W. Principles and applications of ion scattering spectrometry: surface chemical and structural analysis; Wiley: New York, 2003; p 336.
38. Aizawa, M.; Lee, S.; Anderson, S. L. Surf. Sci. 2003, 542, 253– 275
39. Kane, M. D.; Roberts, F. S.; Anderson, S. L. Faraday Discuss. 2013, 162, 323– 340
40. Kane, M. D.; Roberts, F. S.; Anderson, S. L. J. Phys. Chem. C 2015, 119, 1359– 1375
42
41. Kaden, W. E.; Kunkel, W. A.; Roberts, F. S.; Kane, M.; Anderson, S. L. J. Chem. Phys. 2012, 136, 204701– 204712
42. Kaden, W. E.; Kunkel, W. A.; Anderson, S. L. J. Chem. Phys. 2009, 131 (114701) 114701– 114715
43. Kane, M. D.; Roberts, F. S.; Anderson, S. L. Int. J. Mass Spectrom. 2014, 370, 1– 15
44. Chen, P. J.; Goodman, D. W. Surf. Sci. 1994, 312, L767– L773
45. Street, S. C.; Goodman, D. W. Chem. Phys. Solid Surf. 1997, 8, 375– 406
46. Lai, X.; Chusuei, C. C.; Luo, K.; Guo, Q.; Goodman, D. W. Chem. Phys. Lett. 2000, 330, 226– 230
47. Wu, Y.; Garfunkel, E.; Madey, T. E. J. Vac. Sci. Technol., A 1996, 14, 2554– 2563
48. Wu, Y.; Garfunkel, E.; Madey, T. E. Surf. Sci. 1996, 365, 337– 352
49. Chen, M. S.; Goodman, D. W. J. Phys.: Condens. Matter 2008, 20, 264013
50. Kane, M. D.; Roberts, F. S.; Anderson, S. L. Int. J. Mass Spectrom. 2014, 370, 1– 15
51. Kresse, G.; Furthmüller, J. Comput. Mater. Sci. 1996, 6, 15– 50
52. Kresse, G.; Furthmüller, J. Phys. Rev. B: Condens. Matter Mater. Phys. 1996, 54, 11169
53. Kresse, G.; Hafner, J. Phys. Rev. B: Condens. Matter Mater. Phys. 1993, 47, 558
54. Kresse, G.; Hafner, J. Phys. Rev. B: Condens. Matter Mater. Phys. 1994, 49, 14251
55. Kresse, G.; Joubert, D. Phys. Rev. B: Condens. Matter Mater. Phys. 1999, 59, 1758
56. Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. Rev. Lett. 1996, 77, 3865
57. Bourdillon, A.; El-Mashri, S.; Forty, A. Philos. Mag. A 1984, 49, 341– 352
58. Levin, I.; Brandon, D. J. Am. Ceram. Soc. 1998, 81, 1995– 2012
59. Ahn, J.; Rabalais, J. Surf. Sci. 1997, 388, 121– 131
43
60. Wang, X.-G.; Chaka, A.; Scheffler, M. Phys. Rev. Lett. 2000, 84, 3650
61. Zhai, H.; Ha, M.-A.; Alexandrova, A. N. J. Chem. Theory Comput. 2015, 11, 2385– 2393
62. Averkiev, B. Geometry and electronic structure of doped clusters via the Coalescence Kick method. Ph.D. Thesis, Utah State University, December 2009.
63. Tian, W. Q.; Ge, M.; Sahu, B.; Wang, D.; Yamada, T.; Mashiko, S. J. Phys. Chem. A 2004, 108, 3806– 3812
64. Dadras, J.; Jimenez-Izal, E.; Alexandrova, A. N. ACS Catal. 2015, 5, 5719– 5727
65. Ha, M.-A.; Dadras, J.; Alexandrova, A. ACS Catal. 2014, 4, 3570– 3580 44
Table 2.1: Boltzmann populations of adsorbed ethylene of n = 1..3 coverage in the di-σ, sp3 configuration (precursor to dehydrogenation)
Ethylene Coverage n = 1 n = 2 n = 3 En ethylene,glob (eV) -1.29 -1.62 -1.73 ΔQethylene,glob (e) 0.00 -0.02 -0.30 ΣP450K,sp3 7.35% 17.51% 84.26% ΣP700K,sp3 14.93% 29.42% 69.22%
Table 2.2: Adsorbed isomers with C
Cluster Isomer EC (eV) ΣP450K EC (eV) ΣP700K EC (eV) I -7.05 II -8.03 Pt7-C -7.30 -7.38 III -7.70 IV -7.61 I -8.17 II -7.60 Pt8-C III -8.36 -8.16 -8.12 IV -7.90 V -7.27 45
Figure 2.1: The lowest-energy minima of adsorbed Pt and Pt , with adsorption 7 8 energies (E ), Boltzmann population at catalytic temperature of 700 K (P ), and ads 700K charge transfer from the support to the cluster (ΔQ).
46
0.010 C D A) Pt4 2 4 desorption alumina C D desorption - 1st TPD 0.008 2 4 C D desorption - 2nd TPD 2 4 D2 desorption - 1st TPD
0.006 D2 desorption - 2nd TPD Integrated Amounts C2D4 = 0.32 molecules/atom 0.004 D2 = 0.13 molecules/atom C D = 0.32 molecules/atom 2 4 D2 = 0.03 molecules/atom 0.002
0.000 B) Pt7
0.008 Integrated Amounts C D = 0.38 molecules/atom 2 4 0.006 D2 = 0.20 molecules/atom C D = 0.29 molecules/atom 2 4 D2 = 0.07 molecules/atom 0.004
atom/sec Molecules/Pt 0.002
0.000 C) Pt8
0.008 Integrated Amounts C D = 0.31 molecules/atom 0.006 2 4 D = 0.15 molecules/atom 2 C2D4 = 0.30 molecules/atom D = 0.05 molecules/atom 0.004 2
0.002
0.000 100 200 300 400 500 600 700 Temperature (K)
Figure 2.2: Desorption of unreacted C2D4 (solid) and D2 products (circles) during the 1st and 2nd consecutive TPR runs. Each TPR measurement was made after a 5 L C2D4 exposure to Ptn/alumina (n = 4, 7, or 8) at 150 K. The dashed red line shows desorption of unreacted C2D4 from the cluster-free alumina support and is scaled as if the sample contained an identical Pt loading as the Ptn/alumina samples. Each set of spectra results from individual experiments. Included in each frame is the total integrated desorption of C2D4 and D2 per Pt atom for each experiment.
47
Figure 2.3: Structures of ethylene binding to Pt7. Left column: Binding of a single ethylene molecule to different sites on single-layer Pt7. Columns 2 – 4: Binding of 1, 2, or 3 ethylene molecules in different sites on the global minimum prismatic isomer of Pt7. Energetics and bonding analysis are summarized in Table A.5 and adsorption geometries and thermal distributions are summarized in Table 2.2 For additional local minima at each coverage (n = 2, 3), refer to Appendix A. 48
0.008 5 L C 2D4, Pt7 0.1 L C D , Pt 2 4 7 0.01 L C D , Pt 0.006 2 4 7 5 L C D , Alumina Film 2 4 0.1 L C D , Alumina Film 2 4 0.004 Intergrated Amounts Cluster C2D4 = 0.29 molecules/atom C D = 0.18 molecules/atom 2 4 C2D4 = 0.04 molecules/atom 0.002 Alumina Film C2D4 = 0.09 molecules/atom C D = 0.02 molecules/atom 2 4
0.000
Molecules/Pt atom/sec Molecules/Pt Integrated Amounts D (5 L C D ), Pt D = 0.27 molecules/atom 2 2 4 7 2 D 2 (0.1 L C2D4), Pt7 0.006 D2 = 0.23 molecules/atom D 2 (0.01 C2D4), Pt7 D2 = 0.09 molecules/atom
0.004
0.002
0.000 100 200 300 400 500 600 700
Temperature (K)
Figure 2.4: Desorption of unreacted C2D4 (top frame) and D2 products (bottom frame) during TPR after 150 K C2D4 exposure to Pt7/alumina and the cluster-free alumina support. Desorption of C2D4 from the cluster-free alumina support was scaled as if the sample contained an identical Pt loading as the Pt7/alumina samples. Each set of spectra results from individual experiments with either a 5, 0.1, or 0.01 L C2D4 exposure. Each frame also includes the total integrated C2D4 and D2 desorption per Pt atom for each experiment.
49
0.008
CO TPD/cluster free alumina
CO TPD/as-deposited Pt7 0.006 CO TPD/Pt /post 700 K Flash 7 CO TPD/Pt /post 1 C D TPD 7 2 4 CO TPD/Pt7/post 6 C2D4 TPD
0.004
atom/sec CO molelcules/Pt 0.002
0.000 100 200 300 400 500 600 700
Temperature (K)
Figure 2.5: CO desorption from a Pt7/alumina compared with CO desorption from separately prepared Pt7/alumina samples after a 700 K flash, a single C2D4 TPD, and 6 consecutive TPD’s. All samples were exposed to 10 L of CO at 150 K.
50
1400 A) Alumina film
1200 As-deposited Pt7 Al 700 K heat ramp O 1000 Adsorbate-free Pt 7 800 Pt
Counts 600
400
200
0 B) As-deposited Pt7
1200 5 L C2D4 - 130 K Post C D - 1st TPD 1000 2 4 Post C D - 6th TPD 2 4 800
Counts 600
400
200
0 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 E/Eo
Figure 2.6: Raw ISS spectra for Pt7/alumina samples (a) after a 700 K flash and measured as-deposited. Extrapolated as-deposited is represented by the star. (b) Raw ISS spectra Pt7/alumina sample after the sample was exposed to 5 L of C2D4 at a 150 K and cooled to 130 K, after one C2D4 TPD, and after six consecutive C2D4 TPDs.
51
He+ Exposure (mA*Sec)
0 10 20 30 40 0.5 0.008
0.4 0.006
0.3
0.004 Molecules/Pt aom/sec Molecules/Pt
0.2 2
D
Pt:(Al+O) Ratio Pt:(Al+O) and
0.002 4 D
0.1 2 C
0.0 0.000 100200300400500600700 Temperature (K)
Figure 2.7: Pt/(Al + O) ISS intensity ratios for Pt7/alumina after exposure to 5 L of C2D4 at 150 K and during a sequence where the sample was heated to the indicated temperatures (black circles). The as-deposited Pt intensity and calibrated rates of Pt recovery from C2D4 sputtering and for loss of Pt signal due to sputtering are represented by dashed lines. The Pt/(Al + O) intensity ratio (red star) was measured after completion of a single ethylene TPD cycle. 52
Figure 2.8: First-order approximation of coking on Pt7 and Pt8: lowest-energy structures for a single C atom adsorbed on these clusters.
CHAPTER 3
BORON SWITCH FOR SELECTIVITY OF CATALYTIC DEHYDROGENATION ON
SIZE-SELECTED Pt CLUSTERS ON Al2O3
Reprinted with permission from Mai-Anh Ha, Eric Baxter, Ashley C. Cass, Scott L.
Anderson, Anastassia N. Alexandrova, and The Journal of American Chemical Society
139, 11568. Copyright 2017 American Chemical Society
54
3.1 Overview
Size-selected supported clusters of transition metals can be remarkable and highly tunable catalysts. A particular example is Pt clusters deposited on alumina, which have been shown to dehydrogenate hydrocarbons in a size-specific manner. Pt7, of the three sizes studied, is the most active and, therefore, like many other catalysts, deactivates by coking during reactions in hydrocarbon-rich environments. Using a combination of experiment and theory, we show that nanoalloying Pt7 with boron modifies the alkene- binding affinity to reduce coking. From a fundamental perspective, the comparison of experimental and theoretical results shows the importance of considering not simply the most stable cluster isomer, but rather the ensemble of accessible structures as it changes in response to temperature and reagent coverage.
3.2 Introduction
In the subnano-regime of cluster catalysis, size-selected surface-supported clusters often exhibit nonmonotonic trends in reactivity and selectivity, inspiring the hunt for cluster sizes that are particularly active, selective, and resistant to deactivation.1-4 Not only can they exhibit special catalytic properties due to size effects on electronic and geometric structure, but also most or all of the atoms in sub-nanoclusters are available to bind reactants, making them a promising and rising class of catalysts. In addition, size- selected clusters provide a theoretically tractable approach to testing strategies for catalyst improvement. We recently showed that Pt7 deposited on alumina both binds and dehydrogenates ethylene more efficiently than Pt4 or Pt8 on either a per cluster or a per Pt atom basis. This higher activity was shown to result from the diverse cluster
55
morphologies accessible to Pt7, particularly at higher temperatures and reagent coverages.1 However, this finding is bittersweet, because these clusters and especially the most active Pt7 easily deactivate via a combination of coke (i.e., carbon) deposition and sintering. Coke formation deactivates many catalysts in reactions such as Fischer–
Tropsch synthesis,5 cracking of hydrocarbons,6 and alkene dehydrogenation.7 Sintering, that is, cluster migration, ripening, and agglomeration into larger nanoparticles, where fewer atoms are available on the surface, is another major route of activity loss.8 Thus, improved cluster catalysts would sustain the activity and selectivity of the highly promising Ptn, while resisting coking and sintering.
In this work, we test the strategy of nanoalloying to tune the selectivity for dehydrogenation by Ptn/Al2O3, focusing on Pt7, with the goal of minimizing deactivation by coking and sintering. Doping and alloying can be used to tune the properties of bulk
Pt. Alloying Pt with Sn9 and Zn10, 11 has been used for selectivity control and with Pd to reduce sintering.12, 13 Here, our inspiration is drawn from the boration (boron-doping) of extended surfaces of Co and Ni, used in Fischer–Tropsch synthesis and steam methane reforming, respectively.14, 15 Boration of these metal surfaces extended the lifetime of the catalyst by preventing coke adsorption. In general, boron interacting with metals can lead to a variety of interesting phenomena, such as alloy ultrahardening,16 emergence of topological and Kondo insulators,17 exotic magnetism,18 surface reconstructions,19 record coordination chemistry,20 and the selectivity of Pd catalysts in hydrogenation.21, 22
Recently, we began to theoretically probe boron as a dopant for small Pt clusters deposited on magnesia,23 and found it to reduce affinities of these systems to carbon atoms. Building from this promising initial result, we now address the effect of boration 56
on the selectivity of catalytic dehydrogenation and coking sensitivity of Pt7 on alumina, using both ab initio and statistical mechanical theory, in conjunction with experiment. In what follows, we show that nanoalloying with boron dramatically changes the properties toward alkene binding and dehydrogenation.
3.3 Results and Discussion
Size-selected Pt4,7,8 on alumina have been prepared as discussed in detail
1, 24 previously, and then borated by exposure to diborane (B2H6). Boration and its effects on binding and dehydrogenation of a model alkene, ethylene, were probed by temperature-programmed desorption/reaction (TPD/R), low energy ion scattering (ISS), plane wave density-functional theory (PW-DFT) calculations, and molecular dynamics
(MD) simulations. Initial studies suggested that Pt7 is not only the most active, but also the most susceptible to the effect of boration. We therefore focused our experimental and theoretical work on Pt7 and will explore size effects in future studies.
We find that diborane adsorbs dissociatively on the Pt7 clusters, undergoing both
B–H and B–B bond scission, and leaving atoms of boron in the clusters, as it has been
25 26 27 28 29 reported to do also on the surfaces of Ni, Pd, Ru, Fe or steel, Al2O3, and
30 0 Pt/Al2O3. Pt complexes are also well-known for the successful formation of unique boronated complexes containing borenes, boranes, and borylanes.31, 32 Notably,
Söderlund et al. observed the formation of BH3, B3H7, B3H9, B5H9, and B6H10 in fixed bed reactor studies of diborane on Pt/Al2O3, and it is likely that this also occurs in our
30 experiments. ISS of as-deposited Pt7/alumina (Figure 3.1a) shows peaks for O, Al, and
Pt. B, itself, is undetectable due to a combination of low ISS sensitivity for B, low B
57
coverage, and high background at low E/E0. Nonetheless, because adsorbates attenuate
ISS signal from underlying atoms, the presence of diborane and fragments thereof can be inferred by the effects on other signals.
Considering the Pt signal as reported previously,1 efficient substrate-mediated adsorption of background CO (∼5 × 10–11 mbar) leaves ∼0.5 CO molecules adsorbed per cluster, on average, and by extrapolation we estimate that the as-deposited Pt signal is
∼30% below the adsorbate-free limit (indicated by a ★). Only a small recovery of Pt ISS signal is seen after 700 K heating to desorb adventitious CO, indicating that heating also causes structural changes that result in a smaller fraction of Pt in the surface layer.1
Initial exposure of a Pt7/alumina sample to 1.5 L of B2H6 at 130 K results in ∼80% attenuation of the Pt ISS signal (Figure 3.1a), demonstrating that diborane binds
1, 24 efficiently on top of Pt7. Note that 1.5 L exposure may lead to adsorption of more than one diborane per cluster. Because of computational limits, only adsorption of a single diborane adsorbed to Pt7 clusters was pursued (Figure 3.1b).
PW-DFT calculations were performed to probe adsorption of a single diborane on
Pt7 isomers. In these 0 K and in vacuo calculations, binding of diborane on the prismatic global minimum of Pt7 results in B–H bond scission with some hydrogen leaving for Pt sites; on the more catalytically active1 single layer isomer, the B–B bond also breaks with diborane spontaneously decomposing to form BHy fragments. These results are consistent with the large Pt ISS attenuation observed. The Al and O ISS peaks in ISS underwent only a small attenuation upon diborane exposure at 130 K, indicating that only a small amount of diborane binds to alumina at 130 K, possibly at defects, and the Al and O peaks largely recover when the sample is heated to 300 K, indicating that most of this
58 initial coverage desorbs at low temperatures.
In the sample heated to 300 K, the Pt signal recovered to ∼50% of the as- deposited value, indicating some desorption of diborane or its fragments, but with a significant BxHy coverage remaining, attenuating ISS signal from underlying Pt. After heating to 700 K, the Pt signal recovered to the as-deposited value, but was still ∼30% below the expected adsorbate-free limit, and also below the signal observed after heating without diborane exposure.
CO TPD (Figure 3.2) probed the number and energetics of exposed Pt sites. For
CO on as-deposited Pt7, the main desorption peak is between 300 and 600 K, with a small peak below 200 K. If as-deposited Pt7 is first simply heated to 700 K in UHV, the total amount of CO desorbing from Pt sites is reduced by ∼10%, but the temperature dependence is essentially unchanged. A similar effect is observed if the Pt7/alumina is exposed to a saturation dose of D2 and then heated to 700 K (not shown), consistent with the ISS suggesting thermal restructuring causing a small reduction in the number of exposed Pt sites.
Sintering/agglomeration of the Pt7 into larger clusters, with fewer exposed Pt sites, could potentially account for this small decrease in CO desorption; however, a
24 previous study of CO TPD from Ptn/alumina/Re(0001) (2 ≤ n ≤ 18) found that this CO desorption feature increased significantly with increasing cluster size. Therefore, we conclude that the observed decrease in high-temperature CO desorption after heating cannot be explained by sintering/agglomeration alone. As discussed previously,1 theory suggests that the ensemble of Pt7/isomers favors more prismatic isomers that would also provide fewer CO binding sites. 59
In any case, it is clear that boration has a much larger effect. For Pt7 first exposed to 1.5 L of B2H6 and heated to 300 K, the CO desorption is attenuated by ∼40%, and the main CO desorption peak shifts ∼100 K, demonstrating that boration significantly weakens the Pt–CO binding. B2H6 exposure followed by 700 K heating has little additional effect on either the number or the energetics of CO binding sites, despite the observation that 700 K heating results in recovery of the Pt ISS signal to the as-deposited value. The recovery of Pt ISS signal to the as-deposited value following the 700 K heating is further evidence that the observed changes in the CO binding are not a result of thermal sintering.
DFT calculations show that diborane adsorbs dissociatively atop the clusters as fragments of H, BxHy, or BHy (Figure 3.1b), consistent with the low Pt ISS intensity observed after diborane exposure. However, the majority of Pt7B/Al2O3 structures accessible at 700 K feature the boron acting as a B–Osurf anchor between the cluster and the support (R(B–Osurf) ≈ 1.4 Å, isomers I–IV, VII, VIII) with some structures displaying flatter, single-layer geometries with highly coordinated Pt–B bonds (isomers V–VI, IX,
X, Figure 3.1b). All of these structures expose a large fraction of Pt atoms in the surface layer, accounting for high Pt ISS intensity. Pt atoms bonded to B and reduced charge transfer from the support, presumably account for the weakened CO binding. The decomposition of diborane may undergo many pathways,33-35 and a future study will elucidate the complex interactions between the borating agent and size-selected clusters.
1 In our study of nonborated Ptn/alumina, it was shown that the experimental results were consistent with the theoretical finding of cluster size-dependent ensembles of thermally accessible structures. Predicted evolution of the ensembles with respect to both
60
temperature and ethylene binding was essential to interpreting the ISS and ethylene adsorption results. Because of the increase in complexity of the borated system, such detailed experiment–theory comparison is not feasible; however, theory indicates, perhaps not surprisingly, that Pt7B/alumina has an even more complex ensemble than
Pt7/alumina.
In Pt7B, 10 distinct isomers contribute significantly to the ensemble at 700 K, with the global minimum constituting only 59% of the population. For comparison, the global minimum of Pt7 on alumina comprised 66% of the ensemble and for the less active
1 1 Pt8, 88%. Thus, the structural diversity unique to Pt7/alumina leading to a manifold of binding sites is enhanced in the Pt7B/alumina ensemble. Having access to diverse isomers introduces the possibility of at least one of them being dominant in catalysis, making the entire ensemble more active, although in more complicated reactions diversity can have an adverse effect on selectivity. At the same time, Pt7B’s diversity results in a substantial increase in the configurational entropy’s contribution to the free energy of the system
(Appendix B, Tables B.3 and B.4).12 These observations are valid only if all thermodynamically accessible isomers are also kinetically accessible.
The effects of boration on ethylene binding and dehydrogenation on Pt7 were also probed by TPD/R. Figure 3.3b compares the temperature dependence for C2D4 and D2 desorption from separate Pt7/alumina samples, studied as-deposited and after heating to
300 or 700 K, with and without prior 130 K 1.5 L diborane exposure. For as-deposited
Pt7/alumina, unreacted ethylene desorbs in two components. The low-temperature component is identical to that seen for Pt-free alumina and is attributed to ethylene bound to the alumina support. Desorption from Pt7 sites occurs in a broad component from
61
∼200 to 500 K. D2 desorption (dehydrogenation) is not observed for alumina, but appears between ∼300 and 650 K for as-deposited Pt7/alumina. We previously showed that the onset temperature for D2 desorption is determined by the activation energy for C2D4
1 dehydrogenation rather than for D2 desorption. Integrating the desorption signal allows us to estimate the number of C2D4 and D2 molecules desorbing, which, after subtraction of C2D4 desorption from alumina, amounts to ∼2.1 C2D4 and ∼1.5 D2 molecules per Pt7
1 cluster. Assuming that no hydrogen is left on the surface at 700 K, the number of C2D4 molecules initially adsorbed is ∼2.6 per Pt7 cluster. For as-deposited Pt7/alumina heated to 700 K prior to C2D4 TPD/R, the amount of C2D4 (∼1.8/Pt7) and D2 (∼1.3/Pt7) is 10–
15% lower, and shifted to lower temperatures. The reduction in desorption is stronger if
Pt7/alumina is heated to only 300 K prior to C2D4 TPD/R (∼1.4 C2D4, ∼1.0 D2), presumably because 300 K causes some cluster restructuring, but does not desorb adventitious CO.
Crampton et al. recently reported a study of ethylene hydrogenation to ethane over size-selected Ptn deposited on MgO that provides an interesting point of comparison.36 In their experiment, they coadsorbed hydrogen and ethylene before carrying out TPR, and measured desorption of ethane. In our experiments, there could potentially also be hydrogen present on the surface due to dissociative adsorption of ethylene; however, we did not see any evidence for ethane production. This absence of ethane production may simply reflect the relatively low concentration of hydrogen, as compared to a situation where hydrogen is dosed along with ethylene. However, we note that Crampton et al. did not observe hydrogenation for Ptn smaller than Pt10.
The effects of boration, that is, of 1.5 L diborane exposure and heating, are more
62
dramatic. For either 300 or 700 K heating, desorption of C2D4 is strongly attenuated and shifted to lower temperatures (Figure 3.3b). Note that we do observe a small amount of boron deposition on Pt-free alumina films, presumably at defects; however, this is found to have no effect on the amount or temperature of C2D4 desorption. After these contributions are subtracted, the integrated C2D4 desorption is found to be only 0.75/Pt7
(300 K) and 0.9/Pt7 (700 K). Boration has no significant effect on the temperature onset for D2 production, but the amount of D2 is more than 5 times lower than for as-deposited
Pt7/alumina (0.27 and 0.25/Pt7–B2H6 for 300 and 700 K heating, respectively). Assuming again that no hydrogen is left on the surface at 700 K, the total initial coverage of ethylene is ∼0.83/Pt7B and ∼1.02/Pt7B for samples heated to 300 and 700 K, respectively. It is somewhat surprising that there is not a larger difference between the ethylene chemistry on samples prepared by diborane exposure followed by heating to 300 or 700 K. ISS shows that substantially more BxHy adsorbates remain on the surface of the
Pt clusters after 300 K heating; yet they appear to have only a modest effect on the amount of ethylene binding and its propensity to dehydrogenate.
One question is whether any hydrogen is left on the Pt clusters after boration, that is, after diborane exposure and heating to 700 K. If so, this would complicate measurement of ethylene TPD/R because of possible H/D exchange. To test for this process in a somewhat simpler system, we exposed a borated sample to D2, which adsorbs dissociatively on Pt7/alumina, undergoing recombinative desorption between
∼200 and 350 K. If there were significant H concentration on the sample, significant HD
(mass 3) desorption should occur. None was observed, indicating insignificant residual H concentration on the borated samples. 63
From the perspective of coke reduction, these effects of boration should increase the durability of the catalyst. In alkane dehydrogenation, the goal is to produce alkenes plus hydrogen, but to avoid further dehydrogenation to coke precursors like alkylidenes or alkynes.6, 9, 37, 38 It is clear that boration substantially reduces the ethylene adsorption energy to Pt7, such that desorption occurs below the onset temperature for dehydrogenation (Figure 3.3b). This constitutes the main result of the present work.
Boration tempers but does not kill the catalytic activity of Pt clusters and thus provides a lever for adjusting the selectivity of the catalytic process and a way to eventually optimize it.
DFT provides insight into the mechanism for boron’s effects on ethylene binding
– and decomposition. Pt7 on alumina is negatively charged from 1.2 to 1.4 e , depending on cluster isomer, due to electron transfer from alumina. Upon boration, the amount of net electron transfer (ΔQ) to the cluster decreases, ranging from nearly neutral −0.3 to −1 e–, depending on the isomer. Thus, the nucleophilicity of the Pt7B/alumina ensemble is substantially reduced as compared to pure Pt7/alumina. There is charge separation between atoms: positive Pt coordinated to Osurf, negative Pt to Alsurf, and positive B to
Osurf; that is, Pt atoms within the clusters are charged nonuniformly. Negative charge is associated with cluster nucleophilicity and strong ethylene binding;1 thus a substantial reduction in ethylene adsorption energy would be predicted just on the basis of the effects of boron on clusters’ charge, consistent with the TPD/R results.
The propensity for coking is governed by how likely ethylene is to desorb from
1, 7, the catalyst rather than undergo dehydrogenation to form coke precursors (CHy or Cn).
37-41 Therefore, we theoretically probed ethylene binding on the ensemble of 64
Pt7B/alumina structures at relevant temperatures and up to the maximum coverage observed experimentally (∼3 ethylene/Pt7; see Figure 3.3a). Ethylene binds to Pt in either
π- or di-σ-bonded geometries, the latter being associated with further dehydrogenation.1, 7,
37-41 We extracted the structural information for all isomers considered, to construct ensemble percentages of di-σ-bound ethylene as a metric of ethylene activation at rising temperatures and coverages (Figure 3.3a, see the Appendix B for details). As noted above, in the experiment, borated Pt7 binds roughly one ethylene molecule per cluster.
DFT shows that ethylene preferentially binds to the more nucleophilic Pt sites on the cluster periphery and avoids the electropositive B (Appendix B, Figures B.4–B.8, Tables
B.5–B.7). Moreover, with increasing coverage, ethylene reflects less cooperative adsorption on Pt7B than observed on Pt7: it destabilizes the system by ∼0.2 eV/ethylene
1 in Pt7B, but stabilizes by ∼0.3 eV/ethylene in Pt7.
The Pt7B cluster ensemble also activates a decreasing fraction of ethylene as
1 compared to Pt7. Pt7 binds and activates for dehydrogenation more ethylene as the temperature and coverage increase. At 700 K with low coverage, only 15% of Pt7 isomers contain di-σ ethylene; with high coverage, the percentage increases to >69%. On Pt7B the effect is the opposite: as the temperature and coverage increase, less additional ethylene bind in the di-σ fashion. This occurs because the population becomes enriched in the species that do not activate ethylene such as the π-configuration. While the first ethylene may bind in the di-σ fashion, all subsequent ethylene molecules prefer the weaker, π- configuration associated with hydrogenation or the desorption observed in experiment. At higher coverage, configurations containing additional sp3, di-σ bound ethylene drop from
∼90% to 1.7%. This is the key to the reduced activity of Pt7B.
65
Additionally, this illustrates the importance of the ensemble description of cluster catalysts. Note that if we take just the lower limit of temperature and coverage (i.e., consider just global minima with low ethylene content), we would be tempted to conclude that boration promotes rather than suppresses dehydrogenation, based on the prevalence of di-σ-bound ethylene. This result emphasizes that size- and composition- specific properties of surface-deposited cluster catalysts are not just the properties of a single structure, but of the ensembles present under reaction conditions. A number of studies have noted the importance of dynamic fluxionality in the presence of reagents,
42 43 such as gas-phase Au clusters with CO/H2O or H2, H2S splitting by heterotrimetallic anions,44 and supported Au on magnesia45 or ceria46 for CO oxidation. Our theoretical study provides a comprehensive perspective of ensembles, moving the discussion beyond the low-coverage and low-temperature limit into the realm of real catalysis as evidenced by our discussion and agreement with experiment.
Born–Oppenheimer MD simulations reveal further differences between Pt7 and
Pt7B interacting with ethylene (Appendix B, Figures B.9–B.13). In a previous publication, we predicted that in pure Pt7 clusters, prismatic geometries would stabilize to single-layer geometries during ethylene adsorption.1 We observe this in MD trajectories at 450, 700, and 1000 K (Appendix B, Figure B.9). At 450 and 700 K, the prismatic geometry undergoes multiple transformations: it opens from a prism to a distorted hexagon (circa MD step 150) and varies between other prismatic configurations and other single-layer configurations. At 1000 K, ethylene adsorbed to either the prismatic or the single-layer Pt7 converts to the di-σ configuration. On single-layer Pt7, ethylene exhibits
C–H bond activation, followed by H2 formation (Appendix B, Figure B.10). These MD 66
trajectories demonstrate the high reactivity of single-layer structures, particularly, in pure
Pt7.
We considered Pt7B isomers I, II, and V as representatives of different structural classes. At 450 and 700 K with one bound ethylene molecule, the prismatic isomers I and
II undergo flattening into single-layer and extended, branched configurations around the
B–Osurf anchor, exhibiting high fluxionality. At 700 K, the strongly bound, di-σ ethylene either desorbs (on isomer I) or interconverts to π-bound (on isomer II), i.e., reverts to a geometry where it should tend to hydrogenate or desorb. On the single-layer isomer V of
Pt7B, we observe activation of the C–C and C–H bonds. At 450 K, we see the activation of the C–H bonds proceeding to dissociation of 2H. At 700 K, minor restructuring of the single-layer cluster isomer occurs, and the di-σ-bound ethylene converts to π-bound.
Dehydrogenation appears to proceed through a transition of an H atom from C to a neighboring Pt atom. Born–Oppenheimer MD shows that a vast variety of structures of
Pt7B with bound ethylene are dynamically visited at temperatures of 450 and 700 K (see
Appendix B, Figures B.11–B.13). Even though these MD trajectories give only a partial view on the kinetic accessibility of the isomers, the results of dynamics simulations support the high isomeric diversity and the ensemble description used throughout this study. Thus, MD of these representative Pt7B isomers available for catalysis supports the reduced activity of borated Pt7 as compared to pure Pt7.
Evidence that boration stabilized the catalyst is provided by monitoring changes during 6 sequential TPD/R runs (Figure 3.4a). For Pt7/alumina, the amounts of both C2D4 and D2 desorbing decrease substantially, and post-reaction XPS shows significant carbon
1 deposition. For borated Pt7/alumina, there is essentially no change in C2D4 and D2
67
desorption and no XPS-detectable carbon deposition. Indeed, boration lowers the C2D4 binding energy such that desorption, rather than dehydrogenation and coking, is favored.
The CO TPD in Figure 3.4b reveals additional details. For pure Pt7, the number of
CO binding sites is reduced dramatically after 1 C2D4 TPD/R run, and continues to drop during 6 C2D4 TPD/R runs, exhibiting binding site loss from some combination of coke deposition and cluster sintering/restructuring. Borated clusters also show a gradual decrease in the number of CO sites from sequential C2D4 TPD/R runs. Because no C deposits are detected with XPS, we conclude that the effect is likely attributable to cluster sintering, as supported by the smaller calculated binding energies of Pt7B to the support
(Appendix B, Tables B.3 and B.4) and greater fluxionality of shapes seen in MD, as compared to Pt7.
Finally, we consider coking in terms of the affinity of the clusters to C atoms. As a first-order approximation, we analyzed the Boltzmann-weighted ensemble-averages of the C-binding energies (EC) of the isomers that constitute >96% of the population at 450 and 700 K, for Pt7, Pt7B, and Pt8 (Figure 3.4c), the latter included because it tends to have prismatic isomers like Pt7B. Carbon affinity is both temperature- and isomer-dependent, but the temperature dependence for Pt7 and Pt7B is opposite. For Pt7, the ensemble- average EC increases with temperature; that is, the population evolves to include more isomers with high carbon affinity. In contrast, for Pt7B, higher-energy planar configurations (isomers V and VI) in which B is exposed, rather than anchored to alumina (isomers I–IV), exhibit weaker EC. As a result, the evolving ensemble for Pt7B has decreasing carbon affinity and therefore increasing coke-resistance as temperature rises. For Pt8, EC also decreases with temperature, but remains much higher than for Pt7B.
68
Pt7B’s resistance to carbon can again be traced to its reduced negative charge and nucleophilicity.
3.4 Conclusion
We show that nanoalloying of small, alumina-supported Pt clusters with boron has a substantial effect on the selectivity of catalytic dehydrogenation of ethylene.
Boration reduces the ethylene binding energy and thus the tendency toward undesired dehydrogenation to coke precursors. Coking is one of the major mechanisms for cluster catalyst deactivation, and therefore the proposed strategy of its mitigation might be broadly valuable. The effect is linked to cluster morphologies in the statistical ensemble accessible at experimental conditions of temperature and ethylene coverage. As both temperature and coverage increase, borated clusters activate less ethylene for dehydrogenation and bind less carbon more weakly, as an ensemble, while the opposite is true for pure Pt clusters. Fundamentally, this work illustrates how size- and composition- specific properties of cluster catalysts are necessarily ensemble-averages and cannot be described by individual structures, even if they are the global minima.
3.5 Methods
3.5.1 Experimental Section
The experimental protocol has been detailed elsewhere.1 Briefly, experiments were performed using an instrument that allows in situ sample preparation by cluster deposition and characterization by a variety of methods. Pt7/alumina samples were prepared in ultrahigh vacuum (∼1.5 × 10–10 Torr) by growing an alumina thin film (∼3 69
nm) on a Ta(110) single crystal, and soft landing (1 eV/atom) mass-selected Pt7 clusters onto the support. The alumina thin films were grown using procedures adapted from the work of the Goodman47-49 and Madey50, 51 groups. A detailed study by Chen and
Goodman47 concluded that alumina thin (∼1.5 Å) films grown on a Ta(110) single crystal have slightly distorted hexagonal symmetry that can be related to either the (0001) face of
α-Al2O3 or the (111) face of γ-Al2O3. Because the films were inert to a variety of gas molecules under vacuum, the films were proposed to be preferentially oxygen terminated.
In previous publications,52, 53 we demonstrated that the adsorbate binding, reactivity, and electronic properties of Pd clusters deposited on these alumina thin films were independent of film thickness in the 3–10 nm range. In the present study, we used 3–6 nm thick films. Cluster coverage was controlled by monitoring the cluster neutralization
14 2 current. All samples contained Pt7 coverage corresponding to 1.5 × 10 atoms/cm (∼0.1
Pt monolayer). Deposition took ∼5–15 min.
TPD/R measurements were made with a differentially pumped mass spectrometer that views the sample through a ∼2.5 mm diameter orifice in a skimmer cone, which is surrounded by directional dose tubes allowing gas exposures to the sample. For the ethylene TPD/R measurement, the sample was exposed to 5 L of C2D4 at a sample temperature of 150 K. The sample was then cooled to 135 K and ramped to 700 at 3 K/s, while monitoring masses of interest desorbing from the surface. For CO TPD, samples were exposed to 10 L of 13C16O at a sample temperature of 150 K, cooled to 135 K, and ramped to 700 at 3 K/s, while monitoring desorption of 13CO and other masses of interest. Boration was done by exposing samples to 1.5 L of B2H6 at a sample temperature of 130 K, and then ramping the sample temperature up to either 300 or 700
70
at 3 K/s. Note that all experiments were carried out on separately prepared samples to avoid thermal or adsorbate-induced changes to the samples.
Low energy He+ ion scattering spectroscopy (1 keV He+, 45° angle of incidence, normal detection) was used to observe the effects of cluster size, sample heating, boration, and ethylene TPD/R on the fraction of Pt atoms exposed in the surface layer.
ISS peaks result from He+ scattering from single atoms in the surface layer, identifying the masses of those atoms. Because ISS is a destructive technique, all measurements were made on separately prepared samples or at the end of a series of experimental sequences.
3.5.2 Computational
As discussed previously,1 PW-DFT calculations were performed using the Vienna
Ab initio Simulation Package (VASP)54-57 utilizing projector augmented wave potentials58, 59 and the PBE60 functional. A dense Monkhorst–Pack 8 × 8 × 3 k-point grid was implemented for bulk calculations of the α-Al2O3 unit cell with large kinetic energy cutoffs of 520.0 eV. The optimized lattice constants of a = 4.807 Å and c = 13.126 Å exhibited a slight increase of <0.1 Å as compared to experiment, typical of GGA functionals.61, 62 The unit cell was grown to a (3 × 3) surface with the bottom half of the surface kept fixed and a vacuum gap of 15 Å. A 1 × 1 × 1 k-point grid centered at Γ- point, stringent convergence criteria of 10–5 (10–6) eV for geometric (electronic) relaxations, and kinetic energy cutoffs of 400.0 eV were employed in all calculations.
The Adaptive Force Field Coalescence Kick (AFFCK),63 an adaptive global minimum and local minima search based on the Coalescence Kick (CK),64 was used to find gas- phase Pt7B. A per manum search of adsorbed structures consisted of deposition of the
71
lowest 5–6 gas-phase structures (Appendix B, Figure B.1) under PBE levels of theory with a thorough sampling of cluster faces to possible binding sites. It must be noted that with larger gas-phase clusters, the order of the lowest minima may be DFT method dependent.63 All relevant equations such as the adsorption of reagents, sintering penalty,
Gibbs or configurational entropy, among others, may be found in the Appendix B. MD calculations were also performed in VASP requiring electronic iterations to reach a convergence criterion of 10–8 eV per 1 fs time-step. MD trajectories of >1.5 ps were analyzed to compare adsorption behavior of ethylene on pure Pt versus borated Pt. The
Nose–Hoover thermostat was used to equilibrate the system, approximating conditions to that of the NVT ensemble.
3.6 Acknowledgements
This work was supported by the Air Force Office of Scientific Research under a
Basic Research Initiative grant (AFOSR FA9550-16-1-0141) to A.N.A. and S.L.A. M.-
A.H. was also funded by the UCLA Graduate Division Dissertation Year Fellowship.
CPU resources at the DoD (Department of Defense) High Performance Computing
Modernization Program (the U.S. Air Force Research Laboratory DoD Supercomputing
Resource Center–AFRL DSRC, the U.S. Army Engineer Research and Development
Center–ERDC, and the Navy Supercomputing Resource Center–Navy DSRC), Pacific
Northwest National Laboratory’s Environmental Molecular Sciences Laboratory’s
(EMSL) Cascade cluster, Extreme Science and Engineering Discovery Environment’s
(XSEDE) computing resources, and the UCLA-IDRE cluster were used to conduct this work.
72
3.7 References
1. Baxter, E. T.; Ha, M.; Cass, A. C.; Alexandrova, A. N.; Anderson, S. L. ACS Catal. 2017, 7, 3322– 3335
2. Kaden, W. E.; Wu, T.; Kunkel, W. A.; Anderson, S. L. Science 2009, 326, 826– 829
3. Vajda, S.; Pellin, M. J.; Greeley, J. P.; Marshall, C. L.; Curtiss, L. A.; Ballentine, G. A.; Elam, J. W.; Catillon-Mucherie, S.; Redfern, P. C.; Mehmood, F. Nat. Mater. 2009, 8, 213– 216
4. Lee, S.; Molina, L. M.; López, M. J.; Alonso, J. A.; Hammer, B.; Lee, B.; Seifert, S.; Winans, R. E.; Elam, J. W.; Pellin, M. J. Angew. Chem. 2009, 121, 1495– 1499
5. Schulz, H. Appl. Catal., A 1999, 186, 3– 12
6. Rahimi, N.; Karimzadeh, R. Appl. Catal., A 2011, 398, 1– 17
7. Shaikhutdinov, S. K.; Frank, M.; Bäumer, M.; Jackson, S. D.; Oldman, R. J.; Hemminger, J. C.; Freund, H. Catal. Lett. 2002, 80, 115– 122
8. Hansen, T. W.; DeLaRiva, A. T.; Challa, S. R.; Datye, A. K. Acc. Chem. Res. 2013, 46, 1720– 1730
9. Galvita, V.; Siddiqi, G.; Sun, P.; Bell, A. T. J. Catal. 2010, 271, 209– 219
10. Dadras, J.; Shen, L.; Alexandrova, A. J. Phys. Chem. C 2015, 119, 6047– 6055
11. Shen, L.; Dadras, J.; Alexandrova, A. N. Phys. Chem. Chem. Phys. 2014, 16, 2 6436– 26442
12. Ha, M.; Dadras, J.; Alexandrova, A. ACS Catal. 2014, 4, 3570– 3580
13. Johns, T. R.; Gaudet, J. R.; Peterson, E. J.; Miller, J. T.; Stach, E. A.; Kim, C. H.; Balogh, M. P.; Datye, A. K. ChemCatChem 2013, 5, 2636– 2645
14. Tan, K. F.; Chang, J.; Borgna, A.; Saeys, M. J. Catal. 2011, 280, 50– 59
15. Xu, J.; Chen, L.; Tan, K. F.; Borgna, A.; Saeys, M. J. Catal. 2009, 261, 158– 165
16. Levine, J. B.; Tolbert, S. H.; Kaner, R. B. Adv. Funct. Mater. 2009, 19, 3519– 3533
17. Cooley, J.; Aronson, M.; Fisk, Z.; Canfield, P. Phys. Rev. Lett. 1995, 74, 1629 73
18. KüPers, M.; Lutz-Kappelman, L.; Zhang, Y.; Miller, G. J.; Fokwa, B. P. Inorg. Chem. 2016, 55, 5640– 5648
19. Nandula, A.; Trinh, Q. T.; Saeys, M.; Alexandrova, A. N. Angew. Chem., Int. Ed. 2015, 54, 5312– 5316
20. Popov, I. A.; Jian, T.; Lopez, G. V.; Boldyrev, A. I.; Wang, L. S. Nat. Commun. 2015, 6, 8654
21. Chan, C. W. A.; Mahadi, A. H.; Li, M. M.; Corbos, E. C.; Tang, C.; Jones, G.; Kuo, W. C. H.; Cookson, J.; Brown, C. M.; Bishop, P. T. Nat. Commun. 2014, 5, 5787
22. Krawczyk, M.; Sobczak, J.; Palczewska, W. Catal. Lett. 1993, 17, 21– 28
23. Dadras, J.; Jimenez-Izal, E.; Alexandrova, A. N. ACS Catal. 2015, 5, 5719– 5727
24. Roberts, F. S.; Kane, M. D.; Baxter, E. T.; Anderson, S. L. Phys. Chem. Chem. Phys. 2014, 16, 26443– 26457
25. Desrosiers, R. M.; Greve, D. W.; Gellman, A. J. J. Vac. Sci. Technol., A 1997, 15, 2181– 2189
26. Krawczyk, M. Appl. Surf. Sci. 1998, 135, 209– 217
27. Rodriguez, J. A.; Truong, C. M.; Corneille, J.; Goodman, D. W. J. Phys. Chem. 1992, 96, 334– 341
28. Casadesus, P.; Frantz, C.; Gantois, M. Metall. Trans. A 1979, 10, 1739– 1743
29. Weiss, H.; Knight, J.; Shapiro, I. J. Am. Chem. Soc. 1959, 81, 1826– 1827
30. Söderlund, M.; Mäki-Arvela, P.; Eränen, K.; Salmi, T.; Rahkola, R.; Murzin, D. Y. Catal. Lett. 2005, 105, 191– 202
31. Arnold, N.; Braunschweig, H.; Dewhurst, R. D.; Ewing, W. C. J. Am. Chem. Soc. 2016, 138, 76– 79
32. Cui, Q.; Musaev, D. G.; Morokuma, K. Organometallics 1998, 17, 742– 751
33. McKee, M. L. J. Phys. Chem. 1990, 94, 435– 440
34. Curtiss, L. A.; Pople, J. A. J. Chem. Phys. 1989, 90, 2522– 2523
35. Ganguli, P.; McGee, H., Jr. J. Chem. Phys. 1969, 50, 4658– 4660 74
36. Crampton, A. S.; Rötzer, M. D.; Ridge, C. J.; Schweinberger, F. F.; Heiz, U.; Yoon, B.; Landman, U. Nat. Commun. 2016, 7, 10389
37. Neurock, M.; van Santen, R. A. J. Phys. Chem. B 2000, 104, 11127– 11145
38. Chen, Y.; Vlachos, D. G. J. Phys. Chem. C 2010, 114, 4973– 4982
39. Anderson, A. B.; Choe, S. J. Phys. Chem. 1989, 93, 6145– 6149
40. Windham, R. G.; Koel, B. E. J. Phys. Chem. 1990, 94, 1489– 1496
41. Windham, R.; Koel, B. E.; Paffett, M. Langmuir 1988, 4, 1113– 1118
42. Xing, X.; Li, X.; Yoon, B.; Landman, U.; Parks, J. H. Int. J. Mass Spectrom. 2015, 377, 393– 402
43. Gao, M.; Lyalin, A.; Takagi, M.; Maeda, S.; Taketsugu, T. J. Phys. Chem. C 2015, 119, 11120– 11130
44. Adhikari, D.; Raghavachari, K. J. Phys. Chem. A 2016, 120, 466– 472
45. Häkkinen, H.; Abbet, S.; Sanchez, A.; Heiz, U.; Landman, U. Angew. Chem., Int. Ed. 2003, 42, 1297– 1300
46. Ghosh, P.; Farnesi Camellone, M.; Fabris, S. J. Phys. Chem. Lett. 2013, 4, 2256– 2263
47. Chen, P.; Goodman, D. Surf. Sci. 1994, 312, L767–
48. Street, S.; Goodman, D.; King, D.; Woodruff, D. Chem. Phys. Solid Surf. 1997, 8, 375– 406
49. Lai, X.; Chusuei, C. C.; Luo, K.; Guo, Q.; Goodman, D. Chem. Phys. Lett. 2000, 330, 226– 230
50. Wu, Y.; Garfunkel, E.; Madey, T. E. Surf. Sci. 1996, 365, 337– 352
51. Wu, Y.; Garfunkel, E.; Madey, T. E. J. Vac. Sci. Technol., A 1996, 14, 2554– 2563
52. Kane, M. D.; Roberts, F. S.; Anderson, S. L. Faraday Discuss. 2013, 162, 323– 340
53. Kane, M. D.; Roberts, F. S.; Anderson, S. L. J. Phys. Chem. C 2015, 119, 1359– 1375
75
54. Kresse, G.; Hafner, J. Phys. Rev. B: Condens. Matter Mater. Phys. 1993, 47, 558
55. Kresse, G.; Hafner, J. Phys. Rev. B: Condens. Matter Mater. Phys. 1994, 49, 14251
56. Kresse, G.; Furthmüller, J. Comput. Mater. Sci. 1996, 6, 15– 50
57. Kresse, G.; Furthmüller, J. Phys. Rev. B: Condens. Matter Mater. Phys. 1996, 54, 11169
58. Blöchl, P. E. Phys. Rev. B: Condens. Matter Mater. Phys. 1994, 50, 17953
59. Kresse, G.; Joubert, D. Phys. Rev. B: Condens. Matter Mater. Phys. 1999, 59, 1758
60. Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. Rev. Lett. 1996, 77, 3865
61. Bourdillon, A.; El-Mashri, S.; Forty, A. Philos. Mag. A 1984, 49, 341– 352
62. Levin, I.; Brandon, D. J. Am. Ceram. Soc. 1998, 81, 1995– 2012
63. Zhai, H.; Ha, M.; Alexandrova, A. N. J. Chem. Theory Comput. 2015, 11, 2385– 2393
64. Averkiev, B. Ph.D. Thesis, Utah State University, Logan, UT, 2009. 76
Figure 3.1: (a) Raw ISS spectra for Pt /alumina samples (top) measured 7 immediately after depositing 0.1 ML of Pt (blue) and after flashing Pt /alumina 7 7 to 700 K (green). The spectrum for Pt-free alumina is shown for comparison.
(bottom) Raw ISS spectra for as-deposited (blue), after 1.5 L B H exposure at 2 6 130 K (green), after 1.5 L B H at 130 K exposure followed by heating to 300 K 2 6 (red) or 700 K (black), and after 700 K boration followed by 6 C D TPD/R runs 2 4 (gray). The extrapolated value for adsorbate free Pt7/alumina is shown by red ★. (b) Diborane adsorption results in borated Pt sub-nanoclusters (top). The lowest minima of adsorbed isomers of Pt7B with adsorption energy (Eads), adsorption energies of local minima relative to the global minimum (ΔEads), Boltzmann populations at 700 K, and charge transfer (ΔQ) (bottom). Aluminum atoms are dark gray; oxygen, dark red; platinum, light gray; boron, blue; and hydrogen, white.
77
Figure 3.2: Comparison of CO TPD for a set of Pt7/alumina samples that were first exposed to a particular manipulation and then probed by CO TPD (10 L 13CO exposure at 150 K, heating at 3 K/s to 700 K).
78
Figure 3.3: (a) Deposition of ethylene on isomers I, II, and V of Pt7B from DFT. In π-bound ethylene, both C atoms adsorb to a single Pt site and remain sp2- hybridized (bond angles of ∼120° and a C–C bond-length of ∼1.4 Å). Di-σ bound ethylene binds to two Pt sites and becomes sp3-hybridized (bond angles of ∼109° and a C–C bond of ∼1.5 Å). With increasing temperature and coverage, less ethylene binds as di-σ. Additional minima not visualized here may be found in APPENDIX B along with other structural data such as charges and bonding discussion. Aluminum atoms are dark gray; oxygen, dark red; platinum, light gray; boron, blue; carbon, green; and hydrogen, white. (b) Intact C2D4 (solid) and D2 (dots) desorbing from Pt7/alumina samples after various treatments: as-deposited (blue), 300 K flash (black), 700 K flash (green), 1.5 L of B2H6 with 300 K flash (dark red), and 1.5 L of B2H6 with 700 K flash (purple). The (red) dashed line represents ethylene desorption from the cluster-free alumina. 79
Figure 3.4: (a) Intact C2D4 (solid) and D2 (dots) desorbing from Pt7/alumina samples after various treatments: Pt7 1st TPD/R run (green), Pt7 6th C2D4 TPD/R run (dark red), borated Pt7 1st TPD/R run (blue), and borated Pt7 6th TPD/R run (black). (b) CO TPD from Pt7 with different treatments: Pt7 as-deposited (blue), Pt7 after one C2D4 TPD/R run (black), Pt7 after six C2D4 TPD/R runs (green), borated Pt7 as prepared (dark red), borated Pt7 after one C2D4 TPD/R run (gray), and borated Pt7 after six C2D4 TPD/R runs (purple). The number of CO binding sites on pure Pt7/alumina dramatically decreases after dehydrogenation, whereas borated clusters show fewer CO binding sites and shift toward weaker CO binding. (c) First-order approximation of coking: Boltzmann-weighted C-sticking energies for an ensemble. The Pt7B isomers included in the ensemble are visualized at the bottom. As temperature rises, borated Pt exhibits increasing resistance to carbon. Boltzmann weights utilized as-deposited Pt7, Pt8, and Pt7B adsorption energies. Aluminum atoms are dark gray; oxygen, dark red; platinum, light gray; boron, blue; and carbon, green.
CHAPTER 4
DIBORANE INTERACTIONS WITH Pt7/ALUMINA: PREPARATION OF SIZE-
CONTROLLED BORATED Pt MODEL CATALYSTS WITH IMPROVED
COKING RESISTANCE 81
4.1 Overview
Bimetallic catalysts provide the ability to tune catalytic activity, selectivity, and stability. Model catalysts with size-selected bimetallic clusters on well-defined supports offer a useful platform for studying catalytic mechanisms, however, producing size- selected bimetallic clusters can be challenging. In this study, we present a way to prepare bimetallic model (PtnBm/alumina) cluster catalysts by depositing size-selected Pt7 clusters on an alumina thin film, then selectively adding boron by exposure to diborane and heating.
The interactions between Pt7/alumina and diborane were probed using temperature- programmed desorption/reaction (TPD/R), X-ray photoelectron spectroscopy (XPS), low energy ion scattering (ISS), plane wave density functional theory (PW-DFT), and molecular dynamic (MD) simulations. It was found that the diborane exposure/heating process does result in preferential binding of B in association with the Pt clusters. Borated
Pt clusters are of interest because they are known to exhibit reduced affinity to carbon deposition in catalytic dehydrogenation. At high temperatures, theory, in agreement with experiment, shows that boron tends to migrate to sites beneath the Pt clusters forming Pt-
B-Osuf bonds that anchor the clusters to the alumina support.
4.2 Introduction
Carbon deposition (i.e., “coking”) leads to deactivation of catalysts in important reactions such as Fisher-Tropsch synthesis,1 hydrocarbon cracking,2 and alkene dehydrogenation.3 It has been demonstrated that boration of extended surfaces of Co4 and
Ni5 can extend the lifetime of catalysts without compromising their activity toward Fischer-
Tropsch synthesis and steam reforming, respectively. In both processes, coking is the 82 mechanism of deactivation. We recently showed that boration also reduces coking on size- selected Pt clusters deposited on alumina during dehydrogenation of alkenes.6 This type of catalytic system is novel and so-far largely under-investigated, including several aspects of their preparation, experimental characterization, theoretical analysis, and structural, dynamical, and chemical properties.
Model catalysts with atomically size-selected clusters on well characterized supports provide a useful platform for studying catalysis mechanisms, allowing independent control of the size and density of catalytic sites, and facilitating detailed theoretical studies. Bimetallic catalysts provide important opportunities to tune catalytic activity, selectivity, and stability. However, extending the size-selected model catalyst approach to bimetallic clusters is challenging. One approach is to use alloy or dual target cluster sources that directly produce bimetallic clusters in the gas phase, which can then be mass selected and deposited to create bimetallic model catalysts.7-10 This approach is quite general, in principle, however, for several reasons it becomes increasingly difficult as the cluster size increases. The cluster source intensity is “diluted” over an increasing number of possible MxNy combinations, and the intensity is further decreased by the need for high mass-selector resolution to separate closely spaced masses. Intensity is important, because clusters quickly become contaminated due to substrate-mediated adsorption,11-13 even in ultra-high vacuum (UHV). In many cases, natural isotope distributions exacerbate these problems such that clean selection of both size and composition may be impossible except for very small clusters. For example, Pt has major isotopes with atomic masses 194, 195,
196, and 198, and boron has isotopes with atomic masses 10 and 11. Thus, even for clusters containing only three Pt atoms, the width of the Pt isotopologue distribution is greater than 83
the boron mass, resulting in mass overlaps between Pt3Bn and Pt3Bn±1.
One motivation for this paper is to report a complementary approach to producing size-selected bimetallic cluster catalysts, in which mass-selected cluster deposition is used to create a size-selected model catalyst (here, Ptn/alumina), which is then used to seed deposition of a second element to create a bimetallic model catalyst (here, PtnBm/alumina).
The challenge is to find conditions where boron deposits only on the Pt clusters, and then to characterize the nature of the resulting doped clusters.
4.3 Methods
As outlined below, alumina-supported size-selected Pt4,7,8 model catalysts were prepared, and then exposed to diborane and heated to drive decomposition and H2 desorption. The goal is to selectively borate the Pt clusters, thus it is important to understand how diborane interacts with both Ptn clusters and the alumina support. These interactions were probed by temperature-programmed desorption and reaction (TPD/R), low energy He+ ion scattering (ISS), and X-ray photoelectron spectroscopy (XPS) experiments on both Pt-free alumina and Ptn/alumina samples, by plane wave density functional theory (PW-DFT) calculations of adsorption geometries and energetics, and molecular dynamic (MD) simulations of borane surface chemistry. We previously showed that the chemical consequences of boration are similar for different Pt cluster sizes.6 Here we focus on the boration mechanism, using Pt7/alumina as the example system.
84
4.3.1 Computational
As discussed previously in detail,14 PW-DFT calculations with projector augmented wave potentials15-16 and the PBE17 functional were implemented in the Vienna
Ab initio Simulation Package (VASP).18-21 The bulk-optimized unit cell with lattice constants of a = 4.807 Å and c = 13.126 Å was grown to a (3 × 3) surface, a slight expansion as compared to experiment.22-23 A vacuum gap of 15 Å was added to the slab. The bottom half of the surface was kept fixed. For all calculations, convergence criteria of 10-5 (10-6) eV for geometric (electronic) relaxations, expansion of the plane waves’ kinetic energy to
400.0 eV, and a k-point grid of 1 × 1 × 1 centered at the Γ-point were instituted. We
14 previously discussed the global optimization of Pt7 on the model α-alumina surface, finding a number of low-lying isomers for Pt7/alumina. As shown below, the global minimum has Pt7 in a prismatic (i.e., 3D) structure with Pt7-alumina adsorption energy of
-5.09 eV, however, there are isomers only 0.05 eV higher in energy in which all Pt atoms are in a single layer bound to the alumina surface.
A per manum search for diborane adsorption geometries associated with both the prismatic and single layer Pt7/alumina structures was made, starting with the molecule positioned at bridging, hollow, and atomic (atop) sites, oriented both parallel and normal to the surface plane, and rotated in various orientations. The starting geometries focused on adsorption of diborane to the Pt clusters, rather than to the α-Al2O3 surface. The adsorption energy of diborane was calculated via the relation:
EB2H6 = E[Surf + Pt7-B2H6] – E[B2H6]gas – E[Surf + Pt7,glob].
Ab-initio MD calculations, starting at the lowest minimum of diborane adsorbed on both the prismatic and single-layer Pt7/alumina structures, were also performed. 85
Equilibration of the system utilized the Nose-Hoover thermostat and an electronic convergence criterion of 10-8 eV per 1 fs time-step was implemented. The global optimization of Pt4B4 adsorbed on alumina was performed using the Basin Hopping method adapted for surface deposited clusters.24 The local minima search for gas phase
Pt4B4 also utilized Basin Hopping. The adsorption energy of Pt4B4 was taken as:
Eads = E[Surf + Pt4B4] – E[Pt4B4]gas,glob – E[Surf].
4.3.2 Experimental
The experiments were conducted with an instrument consisting of a mass-selected metal cluster ion deposition beamline25 that terminates in an ultrahigh vacuum (~1.5×10-10
Torr) analysis chamber that allows in situ sample preparation and characterization, as discussed previously, along with several of the experimental protocols used here.12, 26-27
The Ptn/alumina model catalysts were prepared on a 7×7 mm Ta(110) single crystal mounted using Ta heating wires to a liquid nitrogen reservoir at the end of a manipulator.
The sample temperature was controlled between 110 and >2100 K by the combination of resistive and electron-bombardment heating and liquid nitrogen cooling. Temperature was measured by a C-type thermocouple spot welded to the back of the Ta single crystal.
Procedures for alumina film growth were adapted from the Goodman28-30 and
Madey31-32 groups. At the beginning of each experiment, the Ta single crystal was annealed above 2100 K for 5 minutes or until no surface contaminants were detected by XPS and
ISS. For alumina film growth, the Ta(110) substrate was transferred to a separately pumped
-6 UHV antechamber, heated to 970 K in 5×10 Torr of O2, while exposed to Al evaporating from a crucible mounted normal to the Ta(110) surface. In previous studies, we 86 demonstrated that the reactivity, adsorbate binding, and electronic properties of Pd clusters deposited on alumina were independent of film thickness in the 3-10 nm range. For these studies the typical growth rate was ~0.2 nm/min and 3-6 nm thick films were used.
Before beginning cluster deposition, the alumina/Ta(110) support was flashed to
~800 K to desorb adventitious adsorbates. To minimize the time the clusters were exposed to background gases, Pt cluster deposition was done as the sample cooled after the flash, beginning when the sample reached ~300 K. The clusters were deposited onto the alumina support through a 2 mm diameter mask, and cluster coverage was controlled by monitoring the neutralization current of soft landed (~1 eV/atom) clusters on the support. Unless stated
13 2 otherwise, all samples were prepared with Pt7 coverage of 2.14 x 10 clusters/cm , amounting to 1.5×1014 Pt atoms/cm2, equivalent to ~10% of a close-packed Pt monolayer.
TPD/R measurements were made using a differentially pumped mass spectrometer that views the main UHV chamber through the ~2.5 mm diameter aperture in a skimmer cone. The cone was surrounded by four directional dose tubes that pointed at the sample position, and gas doses for both TPD/R and diborane exposure were done using the tubes to minimize gas exposures to the vacuum system. To calibrate the exposures, we compared subsaturation CO TPD signals for CO delivered through the dose tubes and through a gas inlet remote from the sample position. During both gas dosing and the subsequent TPD/R heat ramp, the chamber pressure was monitored by a nude ion gauge, and ion signals of interest were measured by the differentially pumped mass spectrometer. The analyte molecules were ionized by electron impact ionization (EI) using electrons with a nominal energy of 65 eV. This energy was found to give the best intensity for D2 and C2D4 detection.
Diborane TPD/R was done by exposing samples to B2H6 at 130 K sample temperature, 87
followed by heating to 700 K. For ethylene TPD/R, the sample was exposed to 5 L of C2D4 at a sample temperature of 150 K (to minimize multilayer adsorption), cooled to 130 K,
+ + and then heated to 700 K while monitoring signals for C2D4 , D2 , and various background gases.
Boron was introduced into the UHV system in the form of a diborane/argon gas mixture that we characterized mass spectrometrically to have actual composition of 4.8% diborane, 85% argon, and 10.2% H2, the latter assumed to result from diborane decomposition during storage.33 Diborane exposures were calculated based on the measured diborane mole fraction. In most experiments, boration was done by exposing the samples to 1.5 L of diborane at a sample temperature of 130 K, followed by heating to
700 K, which was found to be sufficient to drive desorption to completion. Note that 1.5
L diborane exposure corresponds to ~5.8 x 1014 diborane molecules impinging per cm2, that is, smaller than the total number of surface atoms, but almost four times larger than the number of Pt atoms. A few experiments were performed using a 0.5 L diborane exposure, where the number of impinging diborane molecules (1.9 x 1014/cm2) was only
~25% greater than the number of Pt atoms present. The dose variation had little effect on the sample properties, suggesting that 1.5 L should be more than sufficient to saturate the
Pt cluster binding sites.
Because the gas mixture contained hydrogen, and diborane decomposition also produces hydrogen, we studied TPD following pure D2 exposure, in separate experiments.
H2 desorption during diborane TPD/R could not be monitored because the mass 2 background in the mass spectrometer was too high.
XPS (Al Kα) was used to examine both alumina and Pt7/alumina samples after B2H6 88 exposure, both while holding the sample at the 130 K dose temperature, and after heating the sample to 700 K. Since both boron and Pt are present only in the surface layer, and we know the Pt coverage quite precisely, the boron coverage was estimated from the ratio of
B 1s and Pt 4d XPS integrated intensities. The Pt 4d XPS signal was used because of overlap between Pt 4f and Al 2p. Both Pt and B are present at low coverage, and because the B 1s photoemission cross section is ~40 times smaller than that for Pt 4d,34 the boron
XPS signal is quite weak. To improve the signal/noise, the boron XPS measurements were done using samples with double the normal Ptn coverage (i.e., 0.2 ML). Higher cluster coverage undoubtedly resulted in some increase in cluster agglomeration during deposition and heating, however, because the effects of boration do not appear to be very dependent on cluster size,6 a modest degree of agglomeration is unlikely to have a significant effect on the B/Pt ratio. To insure that the diborane exposure was sufficient to saturate the larger number of Pt7 present, we also doubled the diborane dose to 3.0 L.
For ISS, a beam of 1 keV He+ was loosely focused onto the surface at 45° angle of incidence and the energy distribution of He+ scattered along the surface normal was measured. Peaks in ISS result from scattering of He+ from single atoms in the sample, predominantly in the surface layer.35 Multiple scattering and scattering from subsurface layers contributes primarily to a weak background. In these experiments, ISS was used to monitor the intensities associated with Pt, Al, and O atoms in the top sample layer. H is undetectable by ISS, and the boron ISS signal also proved to be undetectable, due to the
+ combination of low He scattering cross section (σscatt ∝ Ztarget), low boron coverage, and rising background in that energy range from multiple scattering. Because ISS is a destructive technique, the ISS experiments were done either on separately prepared 89 samples or at the end of other experimental sequences.
4.4 Results
4.4.1 Temperature Programmed Desorption/Reaction
Following Adsorption of B2H6 and D2
TPD/R experiments were used to identify species desorbing from alumina and
Pt7/alumina surfaces, and the associated temperature dependences (i.e., energetics).
Because the literature shows that diborane can polymerize on surfaces,36 we monitored masses relevant to known boranes of various sizes. Figure C.1 in Appendix C shows the raw TPD signals for ion masses 11, 26, 48, and 59, which are low background masses
+ + + + 11 10 corresponding to B1Hx ,B2Hx , B4Hx , and B5Hx (the B: B isotope ratio is ~80:20).
Figure 4.1 shows the data corrected, as described below, for the estimated contributions from fragmentation of higher boranes during election impact ionization (EI). Data are shown for desorption from both alumina and Pt7/alumina samples, each exposed to 1.5 L of B2H6 at 130 K, then heated at 3 K/sec. Note that desorption starts at ~120 K, i.e., slightly below the dose temperature. This reflects the fact that diborane pumps out of the system slowly, so that there was a small additional exposure as the sample cooled to the
TPD/R start temperature.
The assignment of the ion signals to desorbing species is complicated by the fact that boranes fragment extensively in EI.37 We corrected the TPD signals for fragmentation of diborane, tetraborane (B4H10) and pentaborane (B5H9), which all have tabulated standard
EI mass spectra, but did not attempt to correct for possible contributions from higher boranes (i.e., BnHm, n > 5). It should be noted that ion fragmentation is very sensitive to 90 the internal energy of the neutral. Therefore, analysis of the TPD/R products using standard mass spectra from the NIST database is useful, but my not be quantitative. The dominant
EI fragment ions for boranes tend to preserve the number of boron atoms (BnHm → BnHm-
+ + x ), however, there is some signal for essentially all possible BxHy (x ≤ n, y ≤ m) fragments.37 For example, mass 59 is the strongest peak in the EI mass spectrum of pentaborane (MW = 63.13), and mass 48 is the strongest peak for tetraborane (MW =
53.32), however, pentaborane EI also produces mass 48 with ~18% of the mass 59 intensity. Mass 26 is the most intense peak in the EI mass spectrum of diborane
(MW=27.67), but 26 is also produced at the few percent level by EI of tetra- and pentaborane.
In Figure 4.1, no correction was made to the mass 59 intensity for possible contributions from EI of BnHm (n ≥ 6), but the mass 48 signal was corrected for cracking of pentaborane, the mass 26 signal was corrected for cracking of pentaborane and tetraborane, and the mass 11 signal was corrected for contributions from penta-, tetra-, and diborane, using NIST standard mass spectra.37 As can be seen by comparing Figures 4.1 and C.1, the corrections are generally quite small. The only qualitatively obvious change is that a peak in the raw signal for mass 48 in the 150 – 170 K range is shown to result almost entirely from EI cracking of pentaborane.
One surprise is that the mass 11 signal (11B+ and 10BH+) is quite high, even after subtraction of the expected contributions from EI fragmentation of di-, tetra-, and pentaborane. We considered the possibility that the high mass 11 intensity might be an artifact of high mass spectrometer sensitivity to light masses, but this explanation is ruled out by the excellent agreement of our mass spectrum for diborane with the NIST standard 91 diborane spectrum.37 For example, when we leak diborane into our UHV system, we measure a mass 11:26 intensity ratio of ~0.3:1, in good agreement with the 0.28:1 ratio reported in the NIST database. It is also unlikely that the high mass 11 signal could result from EI fragmentation of higher boranes (BnHm, n ≥ 6), because these, if present in high enough yield to account for such high mass 11 signal, would also result in much higher mass 59 signal than is observed. For example, hexaborane (B6H10) fragments in EI to produce both masses 11 and 59, however, the mass 59 intensity is ~1.5 times that of mass
11.37
Therefore, we conclude that the high mass 11 signal must largely result from desorption of some BHx species, such as borane (BH3). Diborane is a hydrogen-bridge- bonded dimer, with gas-phase dissociation enthalpy to 2 BH3 of only 1.78 eV (i.e., 0.89
38 eV/BH3), and both BH3 and BH2 are detected mass spectrometrically in gas-phase pyrolysis of diborane at 300 °C.39 We observe mass 11 desorption signal at low temperatures, raising the question of how BH2 or BH3 production is energetically feasible.
For reactions of diborane on a surface, the energy required to generate gas-phase BHx may be supplied by recombination reactions (e.g., producing tetra- and pentaborane) or by formation of strong B-surface bonds.
We also looked for possible desorption of diborane surface reaction products during the 130 K dose, by monitoring masses 11, 26, 48, and 59 during diborane dosing. Signals for masses 11 and 26 were observed in a 0.27:1 ratio, as expected for gas phase diborane, indicating that the diborane sticking probability at 130 K is less than unity, and that little or no borane (BH3) desorbs during the dose. Subunit sticking probability is unsurprising, given that 130 K is only 10 K below the peak of the diborane desorption during TPD/R. 92
Small signals were also observed for masses 48 and 59 during the dose, but these were only
1.2 % and 0.4 % of the mass 26 (diborane) signal for the alumina sample and just 0.2 % and 0.1 % of the diborane signal for Pt7/alumina. Clearly, if borane, tetraborane, or pentaborane form on the surface during the diborane dose, they mostly remain adsorbed until the sample is heated.
We estimated desorption energies for the different species by fitting the TPD/R temperature dependence to a second order kinetic model, that is, assuming that the rate- limiting step is recombination of adsorbed BxHy fragments to generate the various boranes observed. The desorption energy distributions are shown in Figures C.2 and C.3, and the desorption energies all fall in the 0.4 to 0.5 eV range.
Diborane surface chemistry will be discussed in more detail after the rest of the experimental and theoretical results are presented. The most important points to keep in mind are:
1. Because all boranes fragment in EI to produce at least some mass 11, the absence
of mass 11 signal above ~200 K implies that desorption of boron-containing species
is complete by 200 K.
2. The mass 11 signal is far too large to be explained by EI fragmentation of BnHm
(n ≥ 2), implying that there is considerable desorption of BHx.
3. The fact that masses 11, 48, and 59 are observed with higher intensities than mass
26 implies that most of the desorbing boron fraction is in the form of reaction
products, rather than diborane.
4. Desorption from alumina and Pt7/alumina are qualitatively similar, as might be
expected, considering that 90% of the Pt7/alumina surface is alumina.
93
5. The total amount of BnHm desorption is ~13% lower when Pt7 is present at 10%
coverage, and the desorption peaks are shifted 5 to 10 K to higher temperatures.
Species such as B4H10 and B5H9 have the H:B ratios that are smaller than that for diborane (3:1). In addition, the XPS results discussed next show that a significant amount of boron remains on the surface after BnHm desorption has gone to completion. It is clear, therefore, that hydrogen must also be desorbing during diborane TPD/R.
Observation of the diborane → H2 desorption channel is not possible, both because the mass 2 background is high, and because the diborane reactant gas mixture has substantial
H2 concentration. To provide some insight into the binding/desorption behavior of hydrogen on Pt7/alumina, we measured D2 desorption from a separate Pt7/alumina sample dosed with 5 L of D2 at 130 K, and the result is shown in Figure C.5. A small amount of
D2 desorption is observed in the temperature range below ~200 K, where borane desorption occurs, but ~90% of hydrogen desorption occurs at higher temperatures, between 200 and 400 K.
4.4.2 X-Ray Photoelectron Spectroscopy
TPD/R probes the BnHm species that desorb upon heating, but from the perspective of selectively borating the Pt clusters, it is more important to understand the fate of the boron that remains on the surface. XPS was used to probe the fraction of B on the samples before and after heating. Figure 4.2 compares B 1s spectra for both alumina and Pt7/alumina samples, after exposure to 3 L of B2H6 at 130 K and after subsequent heating to 700 K. As noted above, the low B 1s photoemission cross section results in poor signal, and the XPS experiments were done using a sample with Pt7 deposited at
94 twice the normal coverage (~3 x 1014 Pt atoms/cm2).
The B 1s XP spectrum of diborane adsorbed at 130 K on Pt7/alumina is noisy
(bottom right frame) but clearly indicates the presence of two components, fit by peaks at
189.7 and 193.9 eV, suggesting the presence of at least two boron chemical environments.
The 130 K spectrum for Pt-free alumina (bottom left) has similar intensity peaking near
193 eV, but the low binding energy intensity is weaker than in the Pt7/alumina sample. A similar two-component fit was used for this spectrum, resulting in peaks centered at 189.7 and 193.1 eV. After heating to 700 K, only a single broad B 1s feature remains for both alumina and Pt7/alumina, peaking at 190.5 eV binding energy.
The integrated intensities, indicated in each frame of the figure, provide additional insight into diborane interactions with alumina and Pt7. Note that the integrated B 1s intensity at 130 K, that is, the total amount of boron adsorbed, is ~45% higher when Pt7 is present. Heating to 700 K to desorb all volatile boron species, leads to loss of ~34% of the initial B 1s signal for alumina, but only ~22% of the (initially larger) B 1s signal for
Pt7/alumina. Thus, after heating, the amount of boron remaining on the sample is ~71% higher when Pt7 is present, even though the Pt7 coverage was only ~20%. This result can be compared to the TPD/R results, which showed ~13% less desorption of boron- containing species when Pt7 was present (at 10% coverage). Taken together, both TPD/R and XPS show that substantially more boron adsorbs when Pt7 is present, but that less desorbs, that is, the presence of a low coverage of Pt7 leads to substantially more boron deposition on the samples. 95
4.4.3 Temperature-Dependent Ion Scattering Spectroscopy
The final experimental probe of diborane-surface interactions was temperature- dependent He+ ion scattering (TD-ISS). TD-ISS involves cooling the sample, exposing it to an adsorbate of interest, then monitoring changes in ISS intensities as the sample is heated. A typical raw ISS scan (Figure C.4) shows distinct peaks for single scattering from
Pt, O, and Al atoms in the surface layer,35 along with a featureless background due to
12, 14 multiple and subsurface scattering. As discussed previously, Pt7 deposits in an ensemble of prismatic and single layer structures, with most of the Pt in the surface layer, and thus detectable by ISS. Adsorbates attenuate ISS signals from the underlying surface through a combination of shadowing, blocking, and reduced ion survival probability.35, 40
For our scattering geometry, attenuation primarily affects signal from atoms directly under, or surrounding, the adsorbate. Thus, adsorbates binding directly on top of the Pt clusters attenuate Pt signal, with little or no effect on Al or O signals. Conversely, adsorption on the alumina film, or at sites around the periphery of the clusters, tends to attenuate Al and
O signals, with little or no effect on the Pt signal. As heating drives desorption, the attenuated ISS signals should tend to recover toward the adsorbate-free values. To the extent that diborane exposure and heating leads to cluster agglomeration, forming larger multilayer Pt particles, this would reduce the fraction of Pt in the surface layer, and thus the Pt ISS signal.
As shown in Figure C.4, there is no obvious ISS signal for boron (E/E0 ≈ 0.26) in the spectrum taken immediately after diborane exposure at 110 K, nor is B ISS signal observed after heating the sample, despite the evidence that boron must be present on these samples (Figures 4.1 and 4.2). Lack of boron signal could be taken as evidence that boron
96 is not in the surface layer, however, boron may simply have been undetectable due to a combination of low coverage, small He+-B scattering cross section (∝ target atomic
35 number ), and high multiple-scattering background at low E/E0.
Figure 4.3 compares the Pt, O, and Al ISS signals as a function of temperature, for a Pt7/alumina sample that was exposed to 1.5 L of B2H6 at 110 K, probed by ISS, and then heated to 700 K in 50 K steps, with an ISS measurement made at each temperature. All spectra were collected with low (0.1 μA) He+ flux impinging at 45° and detected along the surface normal, with 30 second scan time used to minimize sample damage. The horizontal dashed lines show the Pt, O, and Al intensities measured for adsorbate-free Pt7/alumina in a separate experiment. Compared to these adsorbate-free values, the signals measured after the B2H6 dose are attenuated by ~80% for Pt and ~20% for O and Al, indicating that B2H6 binds preferentially in sites that attenuate Pt ISS signal. ISS, thus, is consistent with the
XPS and TPD/R results indicating that diborane binds preferentially in association with
Pt7, and provides the additional insight that some or all of this Pt-associated diborane binds on top of the clusters, where it attenuates scattering from underlying Pt. The ~20% attenuation of Al and O signals indicates that some diborane binds in sites that shadow or block scattering from alumina, which could include both sites around the periphery of the clusters, and on the alumina film remote from the clusters.
Interpretation of changes in signal as the sample is heated requires knowledge of the effects of He+ bombardment occurring during the repeated ISS scans used in TD-ISS.
To probe the rate of Pt loss by sputtering, an experiment was made on a separate adsorbate- free Pt7/alumina sample, held at constant temperature, while taking a series of ISS spectra.
The rate of Pt ISS signal decrease is indicated in Figure 4.3 by the dashed line labeled “Pt 97
signal loss from sputtering.” Conversely, for a diborane-covered Pt7/alumina sample held at 110 K, the Pt signal slowly increased during successive ISS scans, due to sputtering of adsorbates initially bound on top of the clusters, as indicated by the line labeled “Pt signal recovery from B2H6 sputtering.” The Al and O ISS signals were not observed to change significantly in either control experiment, presumably because the diborane coverage on alumina is low, and sputtering of Al or O from the top layer simply exposes more Al and
O in the second layer.
As shown in Figure 4.3, the Pt ISS signal starts to recover significantly faster than would be expected from B2H6 sputtering at ~200 K, gradually recovering to ~95% of the adsorbate-free value by ~550 K, then is constant at higher temperatures. The Al and O signals remain attenuated up to ~450 K but then recover to their adsorbate-free values by
700 K.
4.4.4 DFT Results for Adsorption of Diborane on Pt7 Clusters
PW-DFT calculations were performed to identify low energy adsorption geometries for diborane on Pt7/alumina, as summarized in Figure 4.4. We previously reported on the energetics and geometries of numerous isomers of Pt7 and Pt8 bound to alumina.14 Here, we focus on adsorption of diborane on the two lowest energy minima of
Pt7/alumina, which are shown in the small figures next to the titles of each section of Figure
4.4. The most stable Pt7/alumina structure is prismatic (Eads = -5.09 eV, relative to alumina
+ gas phase Pt7), but there is a single-layer isomer that is only slightly higher in energy
(Eads = -5.04 eV). Seven different isomers of diborane adsorbed on both prismatic and single layer Pt7/alumina are shown, all of which would contribute to the population at 700 98
K and below, according to Boltzmann statistics. In these 0 K, in vacuo calculations, the most stable configurations of diborane on the prismatic Pt7 cluster preserve the B-B bond and are adsorbed atop or peripherally to the cluster. On single-layer Pt7, however, the most stable structures involve B-B bond scission.
There are several factors to bear in mind in comparing theory to experiment. Due to computational limits, our DFT calculations were restricted to adsorption of only a single diborane and focused on the strongest diborane binding sites, that is, diborane binding on the clusters. The experimental diborane coverages were higher and populated binding sites on both Pt7 and the alumina film. As a result, the calculations cannot address complex chemistry such as higher borane formation and desorption. In addition, while the minima found by DFT clearly illustrate a variety of binding arrangements, we cannot guarantee that they represent all possible low energy binding geometries. Indeed, the fluxionality of these clusters is important in their catalysis but also resists facile theoretical description.6,
14, 41-44 The complexity of the problem will certainly increase at elevated temperatures or for increasing coverage of B2H6.
For prismatic Pt7/alumina, the most stable isomer in absence of adsorbates, diborane adsorbs atop or at peripheral sites on the cluster. The B-B bond is preserved with only one or two H atoms transferred from diborane to Pt sites. The B-B bond lengths range from 1.72 to 1.88 Å, compared to 1.76 Å calculated for gas-phase diborane (Figure C.6), in excellent agreement with the experimental value (1.7645 Å).45 Relative to prismatic
Pt7/alumina + gas phase B2H6, the most stable binding geometry for diborane on prismatic
Pt7 has Eads = -1.94 eV, and the other structures shown are all within 0.2 eV (ΔEads). The atomic charges for the various isomers of diborane adsorbed on prismatic Pt7/alumina are
99
shown in Figure C.7.
Diborane binds more strongly to the single layer Pt7/alumina isomer, with Eads = -
3.22 eV for the most stable structure. Note that three of the seven isomers shown involve
B-B bond scission and other isomers feature a B-B bond elongated by 7-14% compared to gas-phase B2H6. All isomers involve transfer of up to three H atoms from B to Pt sites.
The atomic charges for the various isomers of diborane adsorbed on single layer
Pt7/alumina are shown in Figure C.8.
The substantially higher Eads for diborane on the single layer Pt7 isomer implies that with one diborane adsorbed, single layer Pt7/alumina becomes the global minimum by ~1.2 eV. The barrier height for diborane-induced isomerization from the prismatic local minimum to the single layer global minimum is unknown, but comparison with ethylene adsorption is suggestive. DFT also found that ethylene adsorbed more strongly on single layer Pt7/alumina (Eads = -1.97 eV), compared to prismatic Pt7/alumina (Eads = -1.29 eV), and in that case, adsorption of three ethylene molecules was sufficient to eliminate the
14 prismatic-to-single-layer isomerization barrier for Pt7. The difference in adsorption energy for diborane on the two Pt7 isomers is almost twice as large as the difference for ethylene, suggesting that isomerization is not unlikely at the diborane exposures used in the experiments.
4.4.5 Molecular Dynamics Simulations of Diborane/
Pt7/Alumina Thermal Chemistry
To probe adsorbate effects and chemistry at the elevated temperatures used in the experiments, we used Born-Oppenheimer MD simulations to examine the fate of
100
diborane adsorbed on Pt7 at 450 and 700 K. Both of these temperatures are well above the range where BnHm desorption is observed (Figure 4.1) and in the range where H2 desorption occurs on Pt clusters (Figure C.5). This is also the range of interest for ethylene dehydrogenation, which peaks near 450 K for Pt7/alumina and goes to
14 completion below 700 K. The prismatic and single-layer minima of Pt7 represent different initial geometries for diborane to adsorb and react on. Selected highlights from
MD trajectories on each structure at both temperatures are given in Figures 4.5 and 4.6, respectively (each MD time step corresponds to 1 fs).
Starting with the lowest energy minimum for diborane on prismatic Pt7, MD shows that at these elevated temperatures diborane undergoes B-B scission to form BH2, BH3, or
BH4 fragments, which may then re-adsorb onto Pt sites (Figure 4.5). Interestingly, in the
450 K trajectory, diborane dehydrogenated completely with one of the boron atoms moving to a position underneath the Pt7 cluster, forming Pt-B-Osurf bonds and anchoring the cluster to the alumina surface. In contrast, in the 700 K trajectory, the Pt cluster flattened to a triangular, single-layer structure with BH fragments maximizing the number of Pt-B bonds.
Throughout the MD trajectories, at both temperatures, hydrogen atoms are mobile, translating to adsorb onto the Pt cluster, to Al, O atoms on alumina, or forming H2 (shown desorbing from the surface). Moreover, both the 450 and 700 K MD trajectories favored
B-B bond scission early on, within the first 120 fs of the simulation.
Starting with lowest energy isomer of diborane on single-layer Pt7, at either 450 or
700 K, the Pt cluster retains much of its structure with B or BHy fragments making small translations (Figure 4.6). Similar to the prismatic Pt cluster, at 450 K, diborane’s boron atoms either sit on top of the cluster to maximize Pt-B bonds or move below the cluster to 101
form Pt-B-Osurf anchor bonds. At 700 K, BH fragments sit on the Pt cluster facets, forming
3-4 Pt-B bonds.
Due to computational time limitations, we were only able to run a few trajectories, following the dynamics for ~5 ps. Of course, the MD picture is incomplete, and only accesses a small portion of configurational space accessible to our systems. Nonetheless, these MD results give insight into possible decomposition mechanisms of diborane on alumina-supported Pt clusters. One obvious point is that Pt, B, and H atoms are all mobile at these temperatures, consistent with the DFT finding of numerous structures within a few tenths of an eV of the global minimum. By the end of the trajectories, the Pt7 clusters remained intact, but most of the initial B-H bonds had broken, with H atoms binding instead to Pt or to O atoms of the support, and some H atoms recombined to form H2 seen desorbing from the surface, even on the relatively short time scale of the trajectory. Boron atoms prefer to bind either under the cluster, forming Pt-B-Osurf linkages, or to facets of the Pt clusters, forming multiple Pt-B bonds. Boron atoms bound to Pt facets may block preferred carbon adsorption sites or weaken carbon adsorption, which may account for the observed resistance to coking of borated Pt clusters.4-6 However, B bound between the cluster and the support also affects the affinity to C by altering the electronic structure of the system, particularly, the charge transfer from the support to the cluster, as was shown previously.6
At the higher diborane coverages of the experiments, additional processes, presumably, would occur, such as coupling of BH fragments to form higher boranes that might desorb (Figure 4.1) and more extensive recombinative desorption of H2. Over longer time scales, particularly at 700 K, additional hydrogen desorption would almost certainly occur, leaving behind Pt7Bx clusters. The simulations suggest that the Pt7Bx clusters would
102
have a range of Pt morphologies (prismatic and single-layer) and boron binding sites (Pt facets, Pt-B-Osurf), and the cluster structures are likely fluxional at high temperatures.
4.4.6 Pt4B4/Alumina
As a computationally tractable model for PtnBm clusters with higher boron mole fraction, as probably form in the experiments, we chose to study Pt4B4 clusters. The global optimization search for gas-phase and adsorbed Pt4B4 was performed utilizing the Basin
Hopping method,24 and the set of low energy isomers is shown in Figures 4.7 and C.9 for
Pt4B4/alumina and gas-phase Pt4B4, respectively. Both gas-phase and adsorbed Pt4B4 structures tend to be (quasi-)planar. Huynh et al. ascribed the drive towards planarity with increasing boron concentration to the covalent nature of boron-boron bonds and boron- metal bonding in mixed metal-boron clusters in BnAl6-n and LiBnAl6-n systems, and proposed that this effect may be general to other metal-boron systems.4-6, 46 Noticeably, their study found the transition from predominantly 3D to 2D structures occurs when the
Al:B ratio is 1:1. The Pt-B clusters seem to follow a similar pattern, with the single-layer geometry dominating in gas-phase Pt4B4 (isomers i-iii, see Appendix C) as compared to
6 Pt7B (isomers v).
For alumina-supported Pt4B4, the Pt4 moiety is also near-planar, however, the B atoms tend to be bent toward the alumina support to allow formation of short B-Osurf bonds anchoring the Pt4 moiety to the alumina. The B-B bonds in surface-bound clusters are ~1.7
Å, and Pt-B bonds are ~2 Å. The results also show that, even for high boron concentrations, the energetically favorable structures have all the Pt atoms in the surface layer, with most or all of the B atoms underneath the clusters. This has important implications for
103 interpretation of the TD-ISS results, and also means that all the Pt atoms are exposed and available to act as catalytic sites.
In previous publications, we predicted that addition of electropositive boron would temper the highly active and electronegative Pt clusters by reducing the charge of the cluster.6, 47 High electron density favors ethylene adsorption in sp3-hybridized geometries, a precursor to dehydrogenation. On the other hand, sp2-hybridization tends to
1, 3, 48-50 favor hydrogenation or desorption. In pure Pt7 and Pt8 clusters on alumina, Pt atoms were found to have charges ranging from +0.78 e, when bound primarily to Osurf,
14 to -0.73 e, when bound primarily to Alsurf, with net cluster charge being ca. -1 e. For
Pt7B on alumina, the net charge on the clusters dropped to ca. -0.35 e, with strong
6 variations from one isomer to another. With increasing boron content, Pt4B4 clusters on alumina become positively charged, between +0.68 and +1.32 e. In addition, the charge separation between Pt and B atoms increases with increasing boron concentration: Pt remains negative, ranging from -0.5 to -1.1 e, and B atoms are positively charged, between +0.4 e, when forming a mix of Pt-B or B-B bonds within the cluster, and +1.87 e, when forming a B-Osurf anchor. Increasing the B:Pt ratio also increases the stability of supported Pt-B clusters with Pt4B4 adsorbing more strongly by ~1.9 eV as compared to
6 Pt7B (Eads = -4.62 eV ). This may be attributed to the cluster maximizing B-Osurf interactions and optimizing the electrostatic attraction between the electronegative Pt atoms and electropositive B atoms with a 1:1 ratio of Pt:B.51 104
4.5 Discussion
4.5.1 Decomposition of Diborane on Pt7/Alumina
From our previous study of ethylene dehydrogenation on borated Pt clusters, we know that boration of Pt7/alumina substantially reduces the ethylene desorption temperature, resulting in a significant decrease in the fraction of ethylene that undergoes unwanted dehydrogenation.6 The obvious questions are how much boron is deposited on the Pt7/alumina surface by the boration process used, and in what kinds of binding sites is it found.
Figure 4.1 shows that quite complex chemistry occurs when diborane is adsorbed on both alumina and Pt7/alumina surfaces. The chemistry is qualitatively similar for the two surfaces, reflecting the fact that the Pt7 coverage is only 10%. Similarities include the low intensity for diborane desorption (mass 26), and higher intensities for desorption of both BHx (mass 11) and higher boranes such as tetraborane (mass 48) and pentaborane
(mass 59). The fact that desorption is dominated by B1 or Bn (n ≥ 4) species indicates that adsorbed diborane dissociates at low temperatures, undergoing complex recombination chemistry.
The DFT results support this conclusion, showing that, even at 0 K, diborane spontaneously loses H atoms on both prismatic and single layer Pt7, and that on the single layer isomer, B-B bond scission also occurs. Given that the single layer isomer becomes the global minimum upon diborane adsorption, extensive diborane decomposition is expected. That expectation is supported by the MD trajectory results, in which B-B bond scission and Pt7 isomerization is observed on the picosecond time scale at moderate temperatures. Experimentally (Figure C.5) and computationally (Figure 4.6), hydrogen
105 recombinative desorption is observed at moderate temperatures, suggesting that the final state of the borated Pt7/alumina samples consists primarily of Pt and B atoms, binding in some fashion to the alumina support.
The XPS results in Figure 4.2 show that a significant fraction of the boron initially adsorbed as diborane is left behind after thermal desorption is complete. The amount of boron on the surface can be estimated from XPS peak intensities. For this analysis we take advantage of the fact both Pt and B are deposited on the sample surface at low coverage, and that we know the amount of Pt deposited quite precisely (1.5 x 1014 Pt atoms/cm2).
Attenuation by inelastic scattering can be neglected for photoelectrons emitted by atoms in the surface layer, thus for the Pt7/alumina sample, the B/Pt coverage ratio can be calculated from the ratio of integrated B 1s and Pt 4d intensities. The only information needed for this calculation is the ratio of B 1s and Pt 4d photoemission cross sections, for which we
34 -3 used theoretical cross sections reported by Yeh et al. (σB1s = 6.6 x 10 Mb, σPt4d = 2.64 x
10-1 Mb), which we checked against empirical atomic sensitivity factors52 taking the electron attenuation length into account.53 (For our 54.7° x-ray source-analyzer angle, photoemission asymmetry can be neglected).54 For the Pt-free alumina sample, we determined the boron coverage by comparison to the Pt7/alumina sample studied under identical conditions. Because of the extremely weak B 1s signal (Figure 4.2), the resulting boron coverages are estimated to have uncertainties of ±40%.
For the Pt-free alumina film, this analysis gives a boron coverage immediately after
2 2 130 K diborane exposure of ~9.8 B atoms/nm corresponding to ~5 B2H6/nm . From its structure, we can estimate that an intact diborane molecule lying flat on a surface would occupy roughly 0.06 nm2, thus the boron coverage is equivalent to roughly 30% of a close- 106
packed monolayer. That can be compared to the ~20% attenuation of Al and O ISS signals observed after 130 K diborane exposure in the ISS experiment (Figure 4.3). After heating the alumina sample to 700 K, the B 1s signal decreased by ~34% to ~6.5 B/nm2, compared to ~15 O atoms, and 10 Al atoms per nm2 of the alumina film.
For the sample containing 0.2 ML-equivalent of Pt7 clusters, the amount of B2H6
2 2 adsorbed at 130 K increased to ~14.2 B/nm or ~7 B2H6/nm . If we assume the diborane coverage on the alumina portion of the 0.2 ML Pt7/alumina sample is just 0.8 of the coverage observed on Pt-free alumina, we can estimate that ~4 of the B2H6 molecules are on alumina sites, and ~3 are associated with Pt sites. This 0.2 ML-equivalent sample had
2 0.43 Pt7 clusters deposited per nm , leading to the conclusion that ~7 diborane molecules are associated with each Pt7 cluster. The large attenuation of Pt ISS signal upon diborane exposure (Figure 4.3) shows that a significant fraction of the Pt-associated diborane is bound in sites on top of the clusters, but we cannot rule out some diborane in sites around the cluster periphery. We note that for a single B2H6, DFT found that diborane fragments occupy both “on top” and peripheral sites (Figure 4.3). Heating the sample to 700 K resulted in final B coverage of ~11 B/nm2. On Pt-free alumina, the final B coverage was
6.5 B/nm2, thus from the 80:20 alumina:Pt area ratio, we can estimate that of the 11 B/nm2,
~5.2 are bound to alumina sites, and the remaining ~5.8 B are bound to Pt sites, or 13.5 B atoms/Pt7. Prior to heating, there were 7 diborane molecules = 14 B atoms in Pt-associated sites. This observation suggests that diborane initially adsorbed in Pt-associated sites decomposes during heating, leaving nearly all of its boron atoms on the surface. The implication is that essentially all the boranes desorbing from Pt7/alumina (Figure 4.1) can be attributed to diborane initially adsorbed on alumina sites. If that conclusion is correct,
107
we would expect ~10% less borane desorption from 0.1 ML Pt7/alumina, compared to Pt- free alumina, which is reasonably consistent with the observation of ~13% less desorption.
The B 1s binding energies also provide insight into the nature of the binding.
Diborane adsorbed at 130 K on Pt7/alumina gives rise to a high binding energy peak at
193.9 eV and a broader low binding energy feature that peaks around 189.7 eV. For diborane on alumina, there is a peak at 193.1 eV with similar intensity to that for
Pt7/alumina but the signal at low binding energies is much weaker than for the Pt7/alumina.
The higher binding energy features are in the energy range (193 – 193.7 eV) typically reported for fully oxidized boron (B3+) in compounds such as boron oxide or boric acid.55
Elemental boron (B0) is reported to have binding energies around 188 eV,55 thus the broad
189.7 eV features are suggestive of boron in some partially oxidized form, which obviously is more prevalent when Pt7 is present. DFT was used to calculate the charges for a single diborane on both prismatic and single layer isomers of Pt7/alumina, as shown in Figures
C.7 and C.8, respectively. It can be seen that roughly half the boron atoms in the various isomers tend to be fully oxidized (B3+), and half are in intermediate oxidation states (B1.5+ to B1.6+). These results appear to be in good agreement with the observed binding energies.
It should be noted, however, that for the higher diborane coverages in the experiments, higher boranes form on the surface. B 1s binding energies for such species are not known, but we note that an orthocarborane (B10C2H12) film deposited on copper is reported to have
B 1s binding energy of 189.3 eV,56 also in reasonable agreement with the lower binding energy feature. For reference, in previous studies of low temperature diborane adsorption/decomposition on Mo(100) and Ni(100) two B 1s peaks were observed at 189.2 eV and 187.6 eV, but in those experiments the boranes were binding directly to metals, 108
rather than oxides.
After heating to 700 K, both alumina and Pt7/alumina samples show a single broad
B 1s peak at ~190.5 eV, suggesting boron is present in a distribution of intermediate oxidation states. This conclusion is broadly consistent with the distribution of boron
0.5+ 1.9+ oxidation states (B - B ) found for B atoms in Pt-B-Osurf bridge bonds, as shown for
Pt4B4 in Figure 4.7.
In summary, XPS indicates that at 130 K, diborane adsorbs preferentially in association with Pt clusters, compared to the alumina support, and that little, if any, of this
Pt-associated boron desorbs during heating to 700 K. As a result, the boration process investigated leaves Pt clusters with much larger boron coverages than the alumina support.
The final B:Pt ratio for the clusters is estimated to be quite high, but we note that the absolute coverages are uncertain by ~ ±40%, due to the very weak B 1s signal. Note also that both the cluster coverage and diborane exposure used in these XPS experiments was twice those for all the other experiments. It is not clear how these changes might have affected the amount of boron deposited per cluster, however, we did study how the diborane exposure used in boration affected subsequent ethylene TPD/R. Boration with
0.5 L diborane exposure was found to be almost as effective at suppressing dehydrogenation as boration with 1.5 L exposure, that is, at least the chemical effects of boration appear to saturate at exposures below those used in all the experiments described in this report.
XPS also shows that the boration process leads to some boron deposition on the alumina film support, thus it is important to know how the catalytic properties of the samples are affected by the boron atoms on (or in) the support. To address this question,
109
Figure 4.8 compares the ethylene adsorption, desorption, and dehydrogenation behavior of three samples:
1. As-deposited Pt7/alumina with no boron exposure (“Pt7/alumina”).
2. Pt7/alumina borated after cluster deposition, that is, both Pt7 and alumina with
boron (“B/Pt7/alumina”).
3. Pt7 deposited on a pre-borated alumina support, that is, only the alumina was
borated (“Pt7/B/alumina”).
Boration was done using our standard method (1.5 L B2H6 exposure at 130 K, followed by heating to 700 K), and ethylene TPD/R was carried out under conditions
6, 14 identical to those used in our previous studies (5 L C2D4 exposure at 150 K, heating at
3 K/sec to 700 K). Ethylene desorbs from Pt7/alumina in two components. The low temperature feature was shown to result from ethylene desorbing from defect sites in the alumina film, and the broad feature peaking near 300 K results from ethylene adsorbed at
14 Pt7 sites. A substantial amount of D2 desorption is observed above 300 K for
Pt7/alumina, but none is observed for the alumina film alone.
Boration of both the clusters and the support (B/Pt7/alumina) leads to a substantial decrease in the desorption temperature distribution for ethylene, and near-total attenuation of D2 desorption. In contrast, borating only the alumina support (Pt7/B/alumina) has little effect on the Pt7 chemistry. The low temperature ethylene desorption feature attributed to desorption from the alumina support is less intense and sharper for this sample, suggesting that boration of the alumina support weakens the ethylene-alumina binding, possibly due to boron occupying alumina defect sites. Pre-boration has little effect, however, on the amount or temperature dependence of ethylene desorbing from Pt sites, or on the D2 110
production, suggesting that the presence of a small amount of boron in the support has little effect on supported Pt clusters. Clearly, the large effects of boron on chemistry of supported Pt clusters are due to boration of the Pt clusters, rather than the support.
XPS probes boron on the surface, thus providing an indirect method to estimate the fraction that desorbs during heating. TPD/R (Figure 4.1) provides a complementary probe
+ of the desorbing fraction, provided that we can convert the measured BxHy ion signals to fluxes of various neutrals desorbing from the surface. It is impractical to directly calibrate the mass spectrometer for detection of species such as BH3, tetraborane, or pentaborane
(toxic, pyrophoric, not commercially available), but we can do an approximate analysis.
14 We previously calibrated the ion intensity – neutral desorption relationship for C2D4, which should have EI cross section similar to that for B2H6, and also has similar EI cracking behavior. For desorbing BHx, B4H10, and B5H9, we note that the EI cross section should scale roughly with number of B atoms, that is, the total ionization signal should increase roughly linearly with borane size. The extent of EI cracking also increases with borane size,37 however, such that the fraction of the ion signal appearing at the detected mass values decreases with size.6 As a result of these two factors, we expect that the detection sensitivity for the various borane products should be similar, and that they should be similar to that for ethylene. Using this crude approximation, we can estimate that of the ~5 B2H6 molecules found by XPS to be adsorbed at 130 K, the equivalent of ~0.9
2 B2H6/nm (~20%) desorbs in the form of various boranes, which is in reasonable
2 agreement with the ~30% desorption estimated from XPS. For Pt7/alumina, ~7 B2H6/nm
2 adsorb at 130 K, but only the equivalent of ~0.80 B2H6/nm (~11%) desorb, compared to
~21% boron desorption estimated by XPS. Considering the crude approximations
111
required for this analysis, and low XPS signal, the TPD and XPS results are in reasonable agreement regarding the amount of boron lost during heating.
Figures C.2 and C.3 give the desorption energy distributions for the various borane products observed, all of which are below 0.5 eV. The DFT adsorption energies for diborane on Pt7/alumina are all much higher – ranging up to ~3 eV on the single layer isomer. This discrepancy is easily explained. The DFT calculations were done to find the structure and energetics for a single diborane molecule in the strongest binding sites, which are on the Pt clusters. The experiments were done at much higher diborane coverages and include diborane bound to Pt sites and to the alumina film. As shown by XPS, boron bound to the Pt clusters does not desorb during heating, thus the TPD/R experiments are only sensitive to boranes desorbing by recombination of BxHy and H adsorbed on the alumina support, where the binding energies are clearly much lower than for Pt-associated sites.
XPS shows the amount of boron associated with Pt, but provides no insight into the nature of the boron-Pt binding. TPD/R gives the temperature ranges in which boranes
(Figure 4.1) and hydrogen (Figure C.5) desorb, but provides no insight into the sites they desorb from. Analysis of the TD-ISS results (Figure 4.3) in light of the XPS and TPD data provides some of this structural information. The large attenuation of Pt ISS signal, and much smaller attenuations of Al and O signals, are consistent with the XPS results. Both show that diborane adsorbs more efficiently in association with the Pt clusters than on alumina sites, and ISS shows that a significant fraction adsorbs on top of the clusters where it attenuates Pt signal. Figure 4.1 shows that desorption of BnHm species is complete by
~200 K, thus it is surprising that there is no significant recovery of Pt, O, or Al ISS signals
112 as the sample is heated to 200 K. Recovery of the Pt ISS signal occurs in two stages at higher temperatures. Between 200 and 400 K, the Pt signal increases to about half the expected value for adsorbate-free Pt7. This is the temperature range in which hydrogen desorbs from Pt7 (Figure C.5), suggesting that desorption of hydrogen exposed some Pt atoms but that ~half the Pt atoms remained blocked by adsorbed boron. This conclusion is consistent with the XPS results indicating that little of the boron associated with Pt sites desorbs, that is, the borane desorption observed in TPD/R originates almost entirely from the alumina film.
Between 400 and 550 K, the Pt signal recovers to almost the adsorbate-free limit.
Since nothing desorbs in this temperature range, the recovery of Pt ISS signal must reflect a structural change in the Pt clusters, and the DFT results suggest the explanation. Both the MD simulations (Figures 4.5 and 4.6) and the structures found for adsorbed Pt4B4 indicate that the most stable binding sites for boron atoms in PtnBm clusters are in Pt-B-
Osurf bridging sites, where the B atoms are under the Pt cluster, anchoring it to the surface.
As a result, the Pt atoms are in surface layer and detectable by ISS.
The fact that the Pt ISS signal recovers to 95% of the value for adsorbate-free Pt7 also suggests that sintering or agglomeration of the clusters during the boration process is limited, because either process would tend to form larger, 3D clusters in which a smaller fraction of Pt is in the ISS accessible surface layer. Indeed, the final Pt ISS signal is well above what would be expected from He+ sputtering of Pt during the series of ISS scans.
This, too, is consistent with the XPS results, which indicate high diborane coverage on the
Pt clusters, which would tend to shield the underlying Pt from most of the sputtering that occurs for adsorbate-free clusters. 113
4.6 Conclusion
We have shown that diborane exposure followed by heating, preferentially deposits boron atoms on supported Pt clusters, with a much smaller boron coverage on the alumina support. The boron on the alumina support is shown to have essentially no effect on ethylene binding or dehydrogenation on the supported Pt clusters. Therefore, the weakening of the ethylene binding and suppression of dehydrogenation when Ptn/alumina is borated can be attributed to boron atoms associated with the Pt clusters. This boron is found by DFT, in agreement with TD-ISS, to move to sites beneath the Pt clusters, forming
Pt-B-Osurf bonds that anchor the clusters to the support.
4.7 Acknowledgement
This work was supported by the Air Force Office of Scientific Research under a
Basic Research Initiative grant (AFOSR FA9550-16-1-0141) to A.N.A. and S.L.A. M.-
A.H. was also funded by the UCLA Graduate Division Dissertation Year Fellowship.
CPU resources at the DoD (Department of Defense) High Performance Computing
Modernization Program (the US Air Force Research Laboratory DoD Supercomputing
Resource Center—AFRL DSRC, the US Army Engineer Research and Development
Center—ERDC, and the Navy Supercomputing Resource Center—Navy DSRC), Pacific
Northwest National Laboratory’s Environmental Molecular Sciences Laboratory’s
(EMSL) Cascade cluster, Extreme Science and Engineering Discovery Environment’s
(XSEDE) computing resources, and the UCLA-IDRE cluster were used to conduct this work. 114
4.8 References
1. Schulz, H. Short History and Present Trends of Fischer–Tropsch Synthesis. Appl. Catal., A 1999, 186, 3-12.
2. Rahimi, N.; Karimzadeh, R. Catalytic Cracking of Hydrocarbons over Modified Zsm-5 Zeolites to Produce Light Olefins: A Review. Appl. Catal., A 2011, 398, 1-17.
3. Shaikhutdinov, S. K.; Frank, M.; Bäumer, M.; Jackson, S. D.; Oldman, R. J.; Hemminger, J. C.; Freund, H.-J. Effect of Carbon Deposits on Reactivity of Supported Pd Model Catalysts. Catal. Lett. 2002, 80, 115-122.
4. Tan, K. F.; Chang, J.; Borgna, A.; Saeys, M. Effect of Boron Promotion on the Stability of Cobalt Fischer–Tropsch Catalysts. J. Catal. 2011, 280, 50-59.
5. Xu, J.; Chen, L.; Tan, K. F.; Borgna, A.; Saeys, M. Effect of Boron on the Stability of Ni Catalysts During Steam Methane Reforming. J. Catal. 2009, 261, 158-165.
6. Ha, M.-A.; Baxter, E. T.; Cass, A. C.; Anderson, S. L.; Alexandrova, A. N. Boron Switch for Selectivity of Catalytic Dehydrogenation on Size-Selected Pt Clusters on Al2o3. J. Am. Chem. Soc. 2017, 139, 11568-11575.
7. Rousset, J.; Cadrot, A.; Cadete Santos Aires, F.; Renouprez, A.; Mélinon, P.; Perez, A.; Pellarin, M.; Vialle, J.; Broyer, M. Study of Bimetallic Pd–Pt Clusters in Both Free and Supported Phases. J. Chem. Phys. 1995, 102, 8574-8585.
8. Yang, B.; Khadra, G.; Tuaillon-Combes, J.; Tyo, E. C.; Pellin, M. J.; Reinhart, B.; Seifert, S.; Chen, X.; Dupuis, V.; Vajda, S. Temperature-Dependent Evolution of the Oxidation States of Cobalt and Platinum in Co1–Xptx Clusters under H2 and Co + H2 Atmospheres. J. Phys. Chem. C 2016, 120, 21496-21504.
9. Bardotti, L.; Tournus, F.; Albin, C.; Boisron, O.; Dupuis, V. Self-Organisation of Size-Selected Coxpt1-X Clusters on Graphite. Phys. Chem. Chem. Phys. 2014, 16, 26653-26657.
10. Moskovkin, P.; Pisov, S.; Hou, M.; Raufast, C.; Tournus, F.; Favre, L.; Dupuis, V..Mo l Predictions and Experimental Characterization of Co-Pt Alloy Clusters. Eur. Phys. J. D 2007, 43, 27-32.
11. Kaden, W. E.; Kunkel, W. A.; Anderson, S. L. Cluster Size Effects on Sintering, Co Adsorption, and Implantation in Ir/Sio2. J. Chem. Phys. 2009, 131, 114701, 1- 15.
12. Roberts, F. S.; Kane, M. D.; Baxter, E. T.; Anderson, S. L. Oxygen Activation 115
and Co Oxidation over Size-Selected Ptn/Alumina/Re(0001) Model Catalysts: Correlations with Valence Electronic Structure, Physical Structure, and Binding Sites. Phys. Chem. Chem. Phys. 2014, 16, 26443 – 26457.
13. Aizawa, M.; Lee, S.; Anderson, S. L. Deposition Dynamics and Chemical Properties of Size-Selected Ir Clusters on Tio2. Surf. Sci. 2003, 542, 253-275.
14. Baxter, E. T.; Ha, M.-A.; Cass, A. C.; Alexandrova, A. N.; Anderson, S. L. Ethylene Dehydrogenation on Pt4,7,8 Clusters on Al2o3: Strong Cluster Size Dependence Linked to Preferred Catalyst Morphologies. ACS Catal. 2017, 7, 3322-3335.
15. Blöchl, P. E. Projector Augmented-Wave Method. Phys. Rev. B 1994, 50, 17953.
16. Kresse, G.; Joubert, D. From Ultrasoft Pseudopotentials to the Projector Augmented-Wave Method. Phys. Rev. B 1999, 59, 1758.
17. Perdew, J. P.; Burke, K.; Ernzerhof, M. Generalized Gradient Approximation Made Simple. Phys. Rev. lett. 1996, 77, 3865.
18. Kresse, G.; Furthmüller, J. Efficient Iterative Schemes for Ab Initio Total-Energy Calculations Using a Plane-Wave Basis Set. Phys. Rev. B 1996, 54, 11169-86.
19. Kresse, G.; Furthmüller, J. Efficiency of Ab-Initio Total Energy Calculations for Metals and Semiconductors Using a Plane-Wave Basis Set. Comput. Mater. Sci. 1996, 6, 15-50.
20. Kresse, G.; Hafner, J. Ab Initio Molecular Dynamics for Liquid Metals. Phys. Rev. B 1993, 47, 558.
21. Kresse, G.; Hafner, J. Ab Initio Molecular-Dynamics Simulation of the Liquid- Metal–Amorphous-Semiconductor Transition in Germanium. Phys. Rev. B 1994, 49, 14251.
22. Bourdillon, A.; El-Mashri, S.; Forty, A. Application of Tem Extended Electron Energy Loss Fine Structure to the Study of Aluminium Oxide Films. Philos. Mag. A 1984, 49, 341-352.
23. Levin, I.; Brandon, D. Metastable Alumina Polymorphs: Crystal Structures and Transition Sequences. J. Am. Ceram. Soc. 1998, 81, 1995-2012.
24. Wales, D. J.; Doye, J. P. Global Optimization by Basin-Hopping and the Lowest Energy Structures of Lennard-Jones Clusters Containing up to 110 Atoms. J. Phys. Chem. A 1997, 101, 5111-5116.
25. Boyd, K. J.; Lapicki, A.; Aizawa, M.; Anderson, S. L. A Phase-Space- 116
Compressing, Mass-Selecting Beamline for Hyperthermal, Focused Ion Beam Deposition. Rev. Sci. Instrum. 1998, 69, 4106-4115.
26. Kane, M. D.; Roberts, F. S.; Anderson, S. L. Mass-Selected Supported Cluster Catalysts: Size Effects on Co Oxidation Activity, Electronic Structure, and Thermal Stability of Pdn/Alumina (N ≤ 30) Model Catalysts. Int. J. Mass Spectrom. 2014, 370, 1-15.
27. Kane, M. D.; Roberts, F. S.; Anderson, S. L. Effects of Alumina Thickness on Co Oxidation Activity over Pd20/Alumina/Re(0001): Correlated Effects of Alumina Electronic Properties and Pd20 Geometry on Activity. J. Phys. Chem. C 2015, 119, 1359–1375.
28. Chen, P. J.; Goodman, D. W. Epitaxial Growth of Ultrathin Al2o3 Films on Ta(110). Surf. Sci. 1994, 312, L767-L773.
29. Street, S. C.; Goodman, D. W. Chemical and Spectroscopic Studies of Ultrathin Oxide Films. Chem. Phys. Solid Surf. 1997, 8, 375-406.
30. Lai, X.; Chusuei, C. C.; Luo, K.; Guo, Q.; Goodman, D. W. Imaging Ultrathin Al2o3 Films with Scanning Tunneling Microscopy. Chem. Phys. Lett. 2000, 330, 226-230.
31. Wu, Y.; Garfunkel, E.; Madey, T. E. Growth and Oxidation of Ultra-Thin Al Films on the Re(0001) Surface. Surf. Sci. 1996, 365, 337-352.
32. Wu, Y.; Garfunkel, E.; Madey, T. E. Growth of Ultrathin Crystalline Al2o3 Films on Ru(0001) and Re(0001) Surfaces. J. Vac. Sci. Technol., A 1996, 14, 2554- 2563.
33. Yu, J.; Boatz, J. A.; Anderson, S. L. Borane-Aluminum Surface Interactions: Enhanced Fracturing and Generation of Boron-Aluminum Core-Shell Nanoparticles. J. Phys. Chem. C 2017, 121, 14176-14190.
34. Yeh, J. J.; Lindau, I. Atomic Subshell Photoionization Cross Sections and Asymmetry Parameters: 1< Z < 103. At. Data and Nucl. Data Tables 1985, 32, 1- 155.
35. Rabalais, J. W. Principles and Applications of Ion Scattering Spectrometry : Surface Chemical and Structural Analysis; Wiley: New York, 2003; pp 336.
36. Bandiera, J.; Naccache, C.; Mathieu, M. V. Products Formed by the Hydrolysis of Diborane in Contact with Γ-Alumina. C. R. Acad. Sci., Paris, Ser. C 1969, 268, 901-4.
37. Stein, S. E., Ir and Mass Spectra. In Nist Chemistry Webbook, Nist Standard 117
Reference Database Number 69, Mallard, W. G, Linstrom, P. J., Eds; NIST Mass Spec Data Center, National Institute of Standards and Technology: Gaithersburg MD 20899 (http://webbook.nist.gov). 2000.
38. Chase, M. W., Jr. ; Davies, C. A.; Downey, J. R., Jr.; Frurip, D. J.; McDonald, R. A.. anaf Thermochemical Tables, Third Edition. J. Phys. Chem. Ref. Data 1985, 14
39. Fehlner, T. P.; Koski, W. S. Direct Detection of the Borane Molecule and the Boryl Radical by Mass Spectrometry. J. Am. Chem. Soc. 1964, 86, 2733-2734.
40. Kaden, W. E.; Kunkel, W. A.; Roberts, F. S.; Kane, M.; Anderson, S. L. Co Adsorption and Desorption on Size-Selected Pdn/Tio2(110) Model Catalysts: Size Dependence of Binding Sites and Energies, and Support-Mediated Adsorption. J. Chem. Phys. 2012, 136, 204705/1-204705/12.
41. Zhai, H.; Alexandrova, A. N. Fluxionality of Catalytic Clusters: When It Matters and How to Address It. ACS Catal. 2017, 7, 1905-1911.
42. Gao, M.; Lyalin, A.; Takagi, M.; Maeda, S.; Taketsugu, T. Reactivity of Gold Clusters in the Regime of Structural Fluxionality. J. Phys. Chem. C 2015, 119, 11120-11130.
43. Hakkinen, H.; Abbet, S.; Sanchez, A.; Heiz, U.; Landman, U. Structural, Electronic, and Impurity-Doping Effects in Nanoscale Chemistry: Supported Gold Nanoclusters. Angew. Chem., Int. Ed. 2003, 42, 1297-1300.
44. Xing, X.; Li, X.; Yoon, B.; Landman, U.; Parks, J. H. Dynamic Fluxionality and Enhanced Co Adsorption in the Presence of Coadsorbed H 2 O on Free Gold Cluster Cations. Int. J. Mass Spectrom. 2015, 377, 393-402.
45. Sams, R. L.; Blake, T. A.; Sharpe, S. W.; Flaud, J. M.; Lafferty, W. J. High- Resolution Infrared Study of the Ν14, Ν17, and Ν18bands Of11b2h6and10b11bh6. J. Mol. Spectrosc. 1998, 191, 331-342.
46. Huynh, M. T.; Alexandrova, A. N. Persistent Covalency and Planarity in the B N Al6–N 2–and Lib N Al6–N–(N= 0–6) Cluster Ions. J. Phys. Chem. Lett. 2011, 2, 2046-2051.
47. Dadras, J.; Jimenez-Izal, E.; Alexandrova, A. N. Alloying Pt Sub-Nano-Clusters with Boron: Sintering Preventative and Coke Antagonist? ACS Catal. 2015, 5719- 5727.
48. Perry, D. A.; Hemminger, J. C. Σ-Bond Metathesis on a Surface: Dehydrogenation of Cyclohexane on Hydrogen-Saturated Pt (111). J. Am. Chem. Soc. 2000, 122, 8079-8080. 118
49. Carlsson, A.; Madix, R. The Dynamics of Ethylene Adsorption on Pt (111) into Di-Σ and Π-Bonded States. J. Chem. Phys. 2001, 115, 8074-8082.
50. Stöhr, J.; Sette, F.; Johnson, A. L. Near-Edge X-Ray-Absorption Fine-Structure Studies of Chemisorbed Hydrocarbons: Bond Lengths with a Ruler. Phys. Rev. Lett. 1984, 53, 1684.
51. Ha, M.-A.; Dadras, J.; Alexandrova, A. Rutile-Deposited Pt–Pd Clusters: A Hypothesis Regarding the Stability at 50/50 Ratio. ACS Catal. 2014, 4, 3570- 3580.
52. Moulder, J. F.; Stickle, W. F.; Sobol, P. E.; Bomben, K. D.; Chastain, J., King, R.C. J. Handbook of X-Ray Photoelectron Spectroscopy; Physical Electronics: Eden Prarie, MN, 1995.
53. Powell, C. J.; Jablonski, A. Nist Electron Effective-Attenuation-Length Database 1.0; NIST: Gaithersburg, 2001.
54. Briggs, D.; Seah, M. P. Practical Surface Analysis, Volume 1 : Auger and X-Ray Photoelectron Spectroscopy, 2nd ed.; John Wiley & sons: Chichester, 1992; Vol. 1; pp 1-657.
55. NIST X-ray Photoelectron Spectroscopy Database, NIST Standard Reference Database Number 20, National Institute of Standards and Technology, Gaithersburg MD, 20899 (2000), doi:10.18434/T4T88K, (accessed December 20, 2017).
56. Pasquale, F. L.; Kelber, J. A. Site-Specific Electron-Induced Cross-Linking of Ortho-Carborane to Form Semiconducting Boron Carbide. Appl. Surf. Sci. 2012, 258, 2639-2642. 119
12000
m/q = 11 (Pt7/alumina)
m/q = 26 (Pt7/alumina) 10000 m/q = 48 (Pt7/alumina)
m/q = 59 (Pt7/alumina) 8000 m/q = 11 (Pt-free alumina) m/q = 26 (Pt-free alumina) m/q = 48 (Pt-free alumina) m/q = 59 (Pt-free alumina)
6000 Counts Counts
4000
2000
0 100 110 120 130 140 150 160 170 180 190 200 210 220 230 240 250 Temperature (K)
Figure 4.1: TPD spectra for select ion signals, corrected for EI cracking of borane species. Alumina and Pt7/alumina samples were exposed to 1.5 L of diborane at 130 K, then heated at 3 K/sec while monitoring desorption mass spectrometrically. 120
700 B (1s) from Al O Film 700 K B (1s) from Pt /Al O 700 K 600 2 3 7 2 3
500
400
300
200
100
0
-100
-200 Integrated Counts = 1012 Integrated Counts = 1730
600 130 K 130 K
500 XPS Intensity (C.P.S) 400
300
200
100
0
-100
-200 Integrated Counts = 1528 Integrated Counts = 2217 -300 200 195 190 185 200 195 190 185 180
Binding Energy (eV)
Figure 4.2: XPS spectra obtained for both Pt-free alumina (black) and Pt7/alumina (blue) samples following exposure to 3 L of B2H6 at 130 K and after heating to 700 K. 121
1400 Oxygen Aluminum Adsorbate-free Oxygen Signal Platinum 1200
Adsorbate-free Aluminum Signal 1000
800
Pt signal loss from sputtering 600 Adsorbate-free Platinum Signal
Peak Intensity (a.u.) Intensity Peak 400
200 Pt signal recovery from B2H6 sputtering
0 0 100 200 300 400 500 600 700 Temperature (K)
Figure 4.3: TD-ISS of Pt7/alumina exposed to 1.5 K of diborane at 110 K. The intensities for adsorbate-free Pt7/alumina, measured separately, are indicated as horizontal dashed lines. The effects of He+ sputtering on Pt signal in adsorbate-free and diborane-dosed Pt7/alumina held at 110 K are shown as dashed lines labeled “Pt signal loss from sputtering” and “Pt signal recovery from B2H6 sputtering,” respectively.
122
Figure 4.4: The seven lowest minima of diborane (B2H6) adsorbed on the two lowest minima of Pt7, which also represent two different structural classes of Pt clusters, i.e., “prismatic” and “single-layer.” The most stable adsorbate-free Pt7/alumina isomer is prismatic, but with diborane adsorbed the single layer isomer becomes more stable by over 1 eV. Boron atoms are depicted in blue, platinum in light gray, hydrogen in yellow, aluminum in dark gray, oxygen in red.
123
Figure 4.5: MD trajectories of diborane decomposition on prismatic Pt7 reveal that diborane may either split apart to form a B-Osurf anchor or maximize Pt-B bonds by adsorbing onto a Pt cluster facet. The prismatic structure can also distort significantly or form a flattened, single-layer geometry. At 450 K, beyond 3.0 ps, the cluster changed very little with only the hydrogens translating from one atom to the next or H2 diffusing through the vacuum gap. Each MD time step corresponded to 1 fs.
124
Figure 4.6: MD trajectories of the decomposition of diborane on single-layer Pt7 reveal similar bonding trends to prismatic Pt7. The stability of the single-layer structure observed in ground state calculations is retained during MD trajectories at these elevated temperatures of 450 and 700 K. At 450 K, MD steps >3100 resemble MD step 2600. Angled side views of the system at MD steps 3100 and 4028 were taken in order to highlight the B-Osurf anchor. Each MD time step corresponded to 1 fs.
125
Figure 4.7: The three lowest minima of Pt4B4 adsorbed on alumina with their associated adsorption energies (Eads), Boltzmann populations at 700 K (P700K), and Bader charges on individual atoms. Boron atoms are depicted in blue, platinum in light gray, aluminum in dark gray, oxygen in red. Isomer II is very similar to Isomer I but with a B-Osurf anchor broken.
126
0.035
C2D4 - Pt7/alumina 0.030 C2D4 - Pt7/B/alumina
C2D4 - B/Pt7/alumina 0.025
0.020
0.015
molecules/cluster/sec
4
D 0.010
2
C
0.005
0.000
D2 - Pt7/alumina 0.025 D2 - Pt7/B/alumina
D2 - B/Pt7/alumina 0.020
0.015
molecules/cluster/sec
2 0.010
D
0.005
0.000 100 200 300 400 500 600 700 Temperature (K)
Figure 4.8: Thermal desorption spectra of unreacted ethylene and deuterium product obtained from three samples: (Red) As-deposited Pt7/alumina with no boron exposure, (blue) Pt7 deposited on pre-borated alumina, and (black) Pt7/alumina borated after Pt7 deposition. Each sample was exposed to 5 L C2D4 at 150K. Boration was done using our standard method (1.5 L B2H6 at 130 K, heating to 700 K). Separate samples were used for each experiment.
CHAPTER 5
CONCLUSION
128
The work presented in this dissertation has focused on fundamental studies of ethylene desorption and dehydrogenation on alumina-supported size-selected Ptn model catalysts, and the effects of boron doping the clusters. In Chapter 2, we show that ethylene binds to the Ptn clusters with a broad distribution of binding energies. The more strongly bound C2D4 decomposed to produce D2, leading to deactivation of the catalyst through coking. Pt7 was found to be significantly more active than both the deposited Pt4 and Pt8 clusters and also deactivated quicker. To fully understand the experimental results, we found that one must consider the ensemble of accessible cluster isomers and how it evolves with temperature and coverage.
In Chapter 3, we found that boron doping the Ptn clusters shifted the distribution of ethylene binding to lower energies, therefore lowering the affinity for dehydrogenation. To further understand the results of borated clusters, we investigated the interactions between diborane and the Pt7 clusters in Chapter 4. We found that exposing the Pt7/alumina sample to diborane followed by heating resulted in the preferential deposition of boron atoms onto the Pt clusters. Furthermore, we found that that the much smaller boron coverage on the alumina film had no effect on the ethylene binding/dehydrogenation on the Pt clusters. Therefore, we can attribute the decrease in ethylene binding/dehydrogenation to the boration of the Pt clusters. Additionally, the Pt- bound boron was found by DFT, in agreement with TD-ISS, to move to sites beneath the
Pt clusters, forming Pt-B-Osurf bonds that anchor the clusters to the support.
The implication from this study is that using a catalytic reaction to seed deposition of a second element may be a useful strategy to selectively modify a catalyst in a beneficial way. For example, we observed that boron deposited on Pt clusters (by
129
exposure to diborane and heating) reduced the ethylene binding energy, thereby reducing the tendency to undergo dehydrogenation to coke precursors. This approach could be used to coat the fuel lines of supersonic aircrafts with a coke resistant catalytic material that could assist in increasing the available heat sink. The results from this study may be broadly valuable since coking is one of the major mechanisms for deactivation of industrial catalyst.
This study raises a number of opportunities for future research. Dehydrogenation of an alkane to an alkene is the overall goal of the catalyst in hypersonic vehicles and needs to be investigated for the boron doped Pt cluster catalyst. The strategy for producing bimetallic cluster catalysts to tune selectivity could also be investigated for a number of important catalytic materials.
APPENDIX A
SUPPORTING INFORMATION FOR CHAPTER 2
131
A.1 TPD Fitting Method and Results
A distribution of population in sites with different energies for desorption/dehydrogenation, θ(E), is assumed, and then the TPD/R spectra are fit using the first order rate equation:1
−퐸 −푑휃 퐼(푡) 훼 = (휃(퐸) ∙ 휈) 푒푘푇(푡), 푑푡 where I(t) is the desorption as a function of time, ν is a prefactor and T(t) is the temperature as a function of time. θ(E) is adjusted until the simulated I(t) matches the experiment. Because size-selected cluster samples are time consuming to prepare, and irreversibly changed by a single TPD/TPR run, it is simply not practical to extract ν from a series of coverage-dependent experiments on every cluster size. Therefore, the simulations were tested for ν ranging from 1013 to 1015 s-1, covering a range often found in TPD.2 The simulated desorption/dehydrogenation energies shift by only ~7% per order-of-magnitude variation in ν, and in Figure A.3 we present the θ(E) distributions obtained for ν = 1014 s-1.
A.2 ISS Extrapolation Method
Figure S5 shows the normalized Pt ISS intensity as a function of He+ exposure in a sequence of low He+ flux (0.1 μA) ISS measurements. The increase in Pt signal during the initial measurements is evidence of a small coverage of adventitious adsorbates (CO and
H2O as determined by separate TPD measurements) that had adsorbed onto the clusters during the ~ 15 min cluster deposition time. The initial increase in Pt signal is a result of the adsorbates being sputtered off the cluster to expose the underlying Pt to He+ scattering.
The Pt signal eventually reaches a maximum and begins to decrease due to Pt sputtering.
132
To determine the as-deposited value the Pt intensity is extrapolated back to the limit of zero
He+ exposure and zero adsorbate coverage as shown by the fit in Figure A.5.
A.3 Pt4 CO TD-ISS
Figure A.6 shows the results from a 13C16O TPD and a TD-ISS experiment performed in our group by F. Sloan Roberts and Matthew Kane on the
13 16 Pt4/alumina/Re(0001) system. The C O TPD (blue) measurement was collected by
13 16 exposing a freshly prepared Pt4/alumina/Re(0001) sample with 10 L of C O at a 180 K.
The sample was allowed to cool to 140 K before heating the sample at a rate of 3 K/sec while detecting the amount of 13C16O desorbing from the surface.
The TD-ISS measurements were collected by exposing a separately prepared
13 16 Pt4/alumina/Re(0001) sample to 10 L of C O at 180 K and colleting an ISS spectrum with a single low He+ flux (0.1μA) scan. The succeeding points in the TD-ISS curve were collected by heating the sample in 50 K increments and measuring the ISS spectrum at the indicated temperatures. The loss of Pt signal due to sputtering, the recovery of Pt signal due to removal of 13C16O, and the as-deposited Pt signal are represented by sloping dashed lines and were determined on separately prepared samples using the same procedure that was used for Figure 2.2.
From the TPD results, heating the sample to 350 K desorbs ~35 % of the total
13C16O coverage but results in an insignificant recovery in the Pt ISS signal. As the sample is heated from 350 to 650 K there is a sharp recovery of the Pt ISS signal as the remaining
13C16O coverage desorbs. These results suggest that the weakly bound 13C16O that is desorbed at temperatures < 350 K is bound in sites that are inefficient at blocking or
133
shadowing the He+ from scattering off the Pt, such as around the periphery of the cluster.
On the other hand, the 13C16O desorbed at temperatures above 350 K leads to a strong recovery of the Pt ISS signal suggesting it desorbs from sites that efficiently attenuate the signal, such as on top sites.
A.4 Theoretical Methods - Details
The relevant equations regarding formation (Eform), adsorption (Eads), and sintering energies (Es) are described in the following. Eform is VASP’s DFT energy of the gas phase cluster with the component, atomic energies already subtracted. The atomic energies arise from the calculated energies of the elements from which the pseudopotential was generated.
Eads[Ptn] = E[Surf+Ptn]-E[Surf]-Egas,min[Ptn], where E [Surf + Ptn] is the total DFT energy of the supported cluster system, E[Surf] is the total energy of the bare support, and Egas,min[Ptn] is the global minimum of the gas-phase cluster.
An analogous equation to Eads for reagent species (reag) such as ethylene and C (a single carbon atom is used as a first-order approximation to coking) is detailed below:
Ereag = E[Surf+Ptn+reag]-E[Surf+Ptn]-Ereag, where E[Surf+Ptn+reag] is the total DFT energy of the supported cluster system with the reagent species and Ereag is the total energy of the gas-phase cluster. In coverage calculations of ethylene, Ereag will encompass the n × Eethyl, where Eethyl is the energy of ethylene in the gas-phase.
Statistical analysis is performed through use of the Boltzmann probability for i-th
134
−퐸 ⁄푘 푇 configuration (Pi) by taking the Boltzmann distribution of each minimum (푒 푖 퐵 ) divided by the sum of the distributions of all relevant low energy minima:
−퐸 푒 푖/푘퐵푇 Pi = −퐸 푘 푇, ∑ 푒 푖/ 퐵 where Ei is the i-th configuration energy of a gas phase cluster (i.e., Eform as defined above), adsorbed cluster (E[Surf+PtmBn]) or adsorbed cluster with a reagent (E[Surf+Ptn+reag]), kB is the Boltzmann constant, and T is the temperature.
The entropic contribution of relevant minima may also be found by considering the fundamental thermodynamic relation of the Helmholtz free energy (F = U – TS).
Specifically, the Gibbs’ entropy equation (SG) allows us to analyze the effect of discrete states with their respective Boltzmann probabilities on the ensemble of particular cluster types:
SG = -kB∑푖 푃푖 푙푛(푃푖), where the PI are the Boltzmann weights and kB is the Boltzmann constant. In this way, we may analyze the entropic contribution at a catalytically relevant temperature (TSG).
In the gas phase, the septamer and octamer contain many isomers whose energies are within 0.2-0.3 eV of the most stable geometry (Figure A.9).3,4 The gas phase isomers present a mixture of 2D and 3D geometries. Adsorbed structures were formed from the deposition of the lowest 5-6 gas phase structures, with a thorough sampling of cluster faces to possible binding sites. The complexity of the corrugated alumina surface leads to a combination of Pt-Al and Pt-O coordination so that single-layer gas phase isomers crinkle in order to maximize wetting of the surface (observed in Pt7, Isomer II, Main text Fig. 2.1).
As the surface is Al-terminated, Pt coordinating to electropositive Al gains a negative charge. Likewise, Pt-O coordination yields positively charged Pt so that a single cluster
135
features a range of electronic depletion or augmentation from one atom to the next. These atomic charges (Δq) are visualized in the main text’s Fig. 2.1. The charge separation between atoms yields an electrostatic potential that further stabilizes clusters and attenuates their site reactivity.
There is an apparent switch in dimensionality between Pt7 to Pt8, where Pt7 on average features more open geometries that wet the corrugated support. The adsorbed Pt7 clusters feature a prismatic geometry (ΣP700K = 66.67%, Isomers I and IV) and a single- layer geometry (ΣP700K = 33.33%, Isomers II, III, and V). This mix of structures offers a complex and rich set of binding sites for adsorbates. In contrast, all of the isomers of Pt8 are prismatic. In prismatic structures, some of the Pt atoms are buried inside the cluster, becoming unavailable as binding sites. These results are in agreement with the experimental findings that suggest that Pt7 provides more binding sites for ethylene as compared to Pt8. Additionally, there is a greater uniformity in the nature of the exposed binding sites, as can be judged by their partial charges. The septamer optimizes the cluster- support interactions with a relatively high charge transfer (ΔQ > -1.20 e) in the global minimum, and, unsurprisingly, features the most favorable adsorption. The added negative charge does not distribute uniformly over the cluster; instead, there is a polarization of Pt atoms in Pt7. Pt8 preserves the charge transfer behavior of Pt7, but adsorption to the support is weaker: of the global minimum of Pt8, Pt8,glob, it is 0.2 eV weaker than that of Pt7,glob.
The Pt atoms within Pt8,glob are charged more uniformly (details in Table A.4).
136
A.5 References
1. Redhead, P.A. Vacuum, 1962, 12, 203–211
2. Kaden, W.E.; Kunkel, W.A.; Roberts, F.S; Kane, M.; Anderson, S.L. J. Chem. Phys., 2012, 136, 204705
3. Zhai, H.; Ha, M.; Alexandrova, A. N. J. Chem. Theor. Comput. 2015, 11, 2385- 2393.
4. Tian, W. Q.; Ge, M.; Sahu, B.; Wang, D.; Yamada,T.; Mashiko, S. J. Phys. Chem. A 2004, 108, 3806-3812.
137
Table A.1: Formation energies of global minima of Pt7, Pt8
Pt7 Pt8 Eform (eV) -27.05 -31.97 Eform/atom (eV) -3.86 -4.00
Table A.2: Gas phase isomers
Cluster Isomer Δ Eform (eV) P450 K P1000 K I 0.00 75.60% 47.39% II 0.05 19.85% 25.97%
III 0.13 2.63% 10.45% Pt7 IV 0.17 0.99% 6.75% V 0.19 0.60% 5.38% VI 0.21 0.32% 4.06% I 0.00 56.98% 38.31% II 0.03 24.95% 26.42% III 0.06 13.66% 20.15% Pt8 IV 0.10 4.31% 11.99% V 0.26 0.06% 1.81% VI 0.29 0.03% 1.33%
138
Table A3. Gas phase isomers under def2/TZVPP basis with pure and hybrid functionals calculated in TURBOMOLE V6.6 Cluster Isomer Δ Eform Δ Eform Δ Eform Δ Eform Δ Eform (eV), (eV), (eV), (eV), (eV), Multipl- Multipl- Multipl- Multipl- Multipl- icity icity icity icity icity VASP TM/PBE TM/PBE0 TM/TPSS TM/TPSSh Pt7 I 0.00 0.00, 5 0.86, 5 0.34, 5 0.65, 5 II 0.05 0.15, 5* 0.36, 5, I 0.18, 5 0.32, 5, I III 0.13 0.01, 5 0.00, 5 0.00, 5 0.00, 5 IV 0.17 0.20, 5, I 0.72, 5 0.40, 3* 0.71, 3 V 0.19 0.05, 5 0.31, 5, I 0.10, 5 0.15, 5 VI 0.21 0.20, 5* 0.40, 5, I 0.26, 5* 0.16, 5 * geometries had difficulty converging, I = geometries with an imaginary frequency
Table A.4: Adsorbed isomers of Pt7, Pt8 with Boltzmann populations (P) at experimentally relevant temperatures of 450 and 1000 K and charge transfer (ΔQ) from the support
Cluster Isomer Δ Eads (eV) P450 K P700 K ΔQ (e) I 0.00 75.27% 65.89% -1.22 II 0.04 24.49% 32.02% -1.44 III 0.24 0.16% 1.26% -1.33 Pt7 IV 0.27 0.07% 0.77% -1.25 V 0.43 <0.01% 0.05% -1.23 ΣPT Eads (eV) -5.08 -5.07 TSG (eV) 0.010 0.019 I 0.00 97.65% 87.73% -1.24 II 0.17 1.14% 5.03% -1.07 III 0.19 0.74% 3.82% -0.94 IV 0.21 0.41% 2.61% -1.08 Pt8 V 0.30 0.04% 0.57% -1.22 VI 0.39 <0.01% 0.14% -1.08 VII 0.41 <0.01% 0.10% -0.80 ΣPT Eads (eV) -4.89 -4.87 TSG (eV) 0.002 0.009
139
Table A5. Local minima of adsorbed ethylene on Pt7 (Isomer I and II) C-C ΔQet- Bond Con- ΔEethylene Hyb Bond Ads. P450 K P700 K hylene Angles fig. (eV) -rid. Lengt- (e) (°) hs (Å) 115.4– i 0.00 86.93% 72.48% 0.00 sp2 1.41 120.5 100.6– ii 0.10 7.35% 14.81% -0.07 sp3 1.49 114.9 116.0– Pt7, iii 0.11 5.72% 12.59% 0.00 sp2 1.40 120.5 Isomer 99.7– I iv 0.38 <0.01% 0.13% -0.16 sp3 1.49 115.4 97.5– v 0.79 <0.01% <0.01% -0.10 sp3 1.50 114.9 100.6– vi 0.87 <0.01% <0.01% -0.02 sp3 1.51 113.9 115.0– i 0.00 99.64% 96.30% 0.01 sp2 1.42 120.5 114.6– ii 0.24 0.22% 1.90% -0.01 sp2 1.41 121.0 97.5– Pt7, iii 0.26 0.12% 1.26% -0.05 sp3 1.49 115.9 Isomer 114.8– II iv 0.34 0.02% 0.36% -0.03 sp2 1.42 120.4 115.3- v 0.39 <0.01% 0.15% -0.03 sp2 1.42 120.4 100.4– vi 0.49 <0.01% 0.03% -0.05 sp3 1.50 115.5
140
0.010
A) Pt4 C2D4 desorption alumina
0.008 C2D4 desorption - 1st TPD
C2D4 desorption - 2nd TPD
0.006 D2 desorption - 1st TPD
D2 desorption - 2nd TPD 0.004
0.002
0.000 B) Pt7 0.008
0.006
0.004
0.002
atom/second Molecules/Pt
0.000 C) Pt 8 0.008
0.006
0.004
0.002
0.000 100 200 300 400 500 600 700 Temperature (K)
Figure A.1: Intact C2D4 (solid) and D2 (circles) desorbing from Ptn/alumina/Ta(110) (n=4,7,8) sample during two consecutive TPD measurements. Intact C2D4 (red dashed line) desorbing from a cluster free alumina/Ta(110) sample. All samples were exposed to 5 L of C2D4 at 150 K before starting the TPD measurement. The D2 signal has not be corrected for the amount of D2 produced from the fragmentation of C2D4 caused by electron impact ionization.
141
0.004 Deuterium from Ethylene Dehydrogenation Deuterium TPD
0.003
0.002
Molecules
2
D
0.001
0.000 100 200 300 400 500 600 700
Temperature (K)
Figure A.2: D2 desorbing from Pt8/alumina/Ta(110) after exposing the sample to 5 L of D2 at 150 K (red). D2 produced by C2D4 dehydrogenation during a C2D4 TPD/R measurement (black).
142
Figure A.3: Energy of desorption fits for C2D4 and D2 desorbing during the first C2D4 TPD/R experiments for Pt4, Pt7, and Pt8.
143
3500
Pt 3000 7
2500
2000 1500
1000
500 0 Pt 3000 8
Counts/Sec
2500
2000
1500
1000 500
0 340 335 330 325 320 315 310 Binding Energy (eV)
Figure A.4: Pt 4d XPS for Pt7 and Pt8, as-deposited on alumina/Ta(110).
144
0.5
Pt7 as-deposited value Fit through Pt signal 7 0.4
0.3
0.2
Pt/(Al+O) Signal ISS
0.1
0.0 0 100 200 300 400 500 600 700
+ 1 keV He exposure (A * sec)
+ Figure A.5: Normalized Pt intensity for Pt7 as a function of He exposure during a sequence of low He+ flux (0.1 μA) ISS scans. The as-deposited value of the Pt intensity can be determined by extrapolating back to the limit of zero exposure and adsorbate coverage as shown by the fit.
145
0.7 4000
As-Deposited Pt4 intensity 0.6 Calibrated Pt4 removal from sputtering 3000 0.5
0.4 2000
0.3 Counts
signal ISS Pt/(Al+O) 0.2 1000 recovery from CO sputtering 0.1 4 Calibrated Pt
0.0 0 100 200 300 400 500 600 700
Temperature (K)
Figure A.6: Pt/(Al+O) ISS signal as function of temperature (circles) after exposing the sample with 10 L of 13C16O at 180 K. CO desorbing from a separately prepared 13 16 Pt4/alumina/Re(0001) sample exposed to 10 L of C O at 180 K during the first TPD measurement (blue). Both samples contained a 0.1 ML of deposited Pt4 clusters.
146
3.5
3.0
2.5
cluster
7 2.0
1.5
molecules/Pt
4
D 1.0 2 C 0.5 C D desorbing from sites on the alumina film 2 4
0.0
1.4
1.2
cluster cluster 1.0 7 0.8
0.6
molecules/Pt 2 D 0.4
0.2
0.0 1st 2nd 3rd 4th 5th 6th
TPD Cycle
Figure A.7: Integrated amounts of C2D4 (top) and D2 (bottom) desorbing, per deposited Pt7 cluster, during a sequence of 6 TPD/R runs under the conditions used in Fig. 2.2 The dashed horizontal line gives an estimate for the C2D4 desorbing from alumina sites, taken from the integrated desorption measured for cluster-free alumina.
147
2000 2000 C 1s raw data Pt 4d raw data C 1s fit Pt 4d fit 1500 1500
1000 1000
counts/sec counts/sec 500 500
0 0 295 290 285 280 340 330 320 310 300 290 Binding Energy (eV)
Figure A.8: C 1s and Pt 4d peaks measured by XPS after 6 consecutive TPD cycles on a Pt7/alumina sample. The presence of carbon suggest that dehydrogenation of ethylene leads to the deposition of carbon onto the model catalyst.
148
Figure A9: Gas phase isomers of Pt7, Pt8 at catalytically relevant temperature of 1000K.
149
Figure A.10: The lowest minima of ethylene adsorbed on Pt7, glob, with calculated Bader charges.
150
Figure A.11: The lowest minima of ethylene adsorbed on the second lowest minimum of adsorbed Pt7, with calculated Bader charges.
151
Figure A.12: The lowest minima of 2 ethylene adsorbed on the Pt7, Isomer I (the global minimum).
152
Figure A.13: The lowest minima of 3 ethylene adsorbed on the Pt7, isomer I (the global minimum).
153
Figure A:14: The C-sticking configurations of the lowest four isomers of Pt7, with calculated Bader charges.
154
Figure A.15: The C-sticking configurations of the lowest five isomers of Pt8, with calculated Bader charges.
APPENDIX B
SUPPORTING INFORMATION FOR CHAPTER 3 156
B.1 Experimental discussion
Further discussion of Fig. 3.2 in the main text: A small amount of CO is seen desorbing from the alumina film at low temperatures, presumably from defect sites in the alumina film. However, strong bimodal desorption of CO from the as-deposited
Pt7/alumina is seen, which we attribute to CO desorbing from sites related to the Pt cluster. Heating the sample to 700 K results in a small attenuation in high temperature desorption component and a small increase in the low temperature desorption component.
Exposing the sample to 1.5 L of B2H6 and heating it to 300 K prior to the CO TPD resulted in the amount of CO desorbing at temperatures above ~350 K to decrease by
70%, presumably from a combination of cluster restructuring and site blocking from residual boron species left on the cluster. Heating the sample to 700 K following the 1.5
B2H6 exposure resulted in a small loss of CO desorption in the 300-400 K range. Further discussion of Fig. 3.3 in the main text: Simply heating Pt7/alumina to 300 K results in a
~38% decrease in the amount of C2D4 desorbing from Pt sites above 200 K (1.3
C2D4/Pt7), and also a ~28% reduction in the amount D2 production (1.0 D2/Pt7).
Heating to 700 K has no further effect on C2D4 desorption above ~300 K, but results in a significant increase in desorption at lower temperatures, such that the total C2D4 desorption (1.9/Pt7) is only ~10% below that for as-deposited Pt7/alumina. D2 production (1.3/Pt7) recovers nearly to the as-deposited value, but shifts to slightly lower temperatures. These effects are attributed to a combination of thermal changes to the cluster morphology, and desorption at temperatures above 300 K, of the small amount of adventitious CO present on the as-deposited Pt7,1 consistent with the ISS results in Figure
3.1. 157
B.2 Additional Theoretical Methods
The relevant equations regarding formation (Eform), adsorption (Eads), and sintering energies (Es) are described in the following. Eform is VASP’s DFT energy of the gas phase cluster with the component, atomic energies already subtracted. The atomic energies arise from the calculated energies of the elements from which the pseudopotential was generated.
퐸푎푑푠[Pt7B] = 퐸[Surf + Pt7B] − 퐸[Surf] − 퐸𝑔푎푠,푚푖푛[Pt7B], where E [Surf +Pt7B] is the total DFT energy of the supported cluster system, E[Surf] is the total energy of the bare support, and Egas,min[PtmBn] is the global minimum of the gas-phase cluster. Table
B.3 lists the adsorption energies of the global minima of adsorbed clusters as well as the sintering energy penalty, that is, the energy cost of an atom of a given element to break away from an octomer, forming a septamer and a monomer on the support:
퐸푠[푃푡7퐵 − 퐵] = 퐸[푆푢푟푓 + 푃푡7] + 퐸[푆푢푟푓 + 퐵1] − 퐸[푆푢푟푓 + 푃푡7퐵] − 퐸[푆푢푟푓]
퐸푠[푃푡8 − 푃푡] = 퐸[푆푢푟푓 + 푃푡8] + 퐸[푆푢푟푓 + 푃푡1] − 퐸[푆푢푟푓 + 푃푡7] − 퐸[푆푢푟푓].
In Table B.3, the sintering energy penalty refers to the monomer energy 퐸[Surf + Pt/B1] in the most favorable position on the support from potential energy surface (PES) calculations. Our PES utilized a fine 10 × 10 grid on a sample unit of our (3 × 3) surface.
An analogous equation to 퐸푎푑푠 for reagent species (reag) such as diborane, ethylene, and
C (a single carbon atom is used as a first-order approximation to coking) is detailed below:
퐸푟푒푎𝑔 = 퐸[Surf + Pt7B + 푟푒푎푔] − 퐸[Surf + Pt7B] − 퐸푟푒푎𝑔 where 퐸[Surf + Pt푚B푛 + 푟푒푎푔] is the total DFT energy of the supported cluster system with the reagent species and 퐸푟푒푎𝑔 is the total energy of the gas-phase cluster. In coverage calculations of ethylene, ethylene adsorption reflected our method of sequential adsorption:
퐸2 푒푡ℎ푦푙푒푛푒 = 퐸2 푒푡ℎ푦푙푒푛푒+𝑔푙표푏,푃푡7퐵푎푑푠 − 퐸1 푒푡ℎ푦푙푒푛푒+𝑔푙표푏,푃푡7퐵푎푑푠 − 퐸1 푒푡ℎ푦푙푒푛푒,𝑔푎푠 158
퐸3 푒푡ℎ푦푙푒푛푒 = 퐸3 푒푡ℎ푦푙푒푛푒+𝑔푙표푏,푃푡7퐵푎푑푠 − 퐸2 푒푡ℎ푦푙푒푛푒+𝑔푙표푏,푃푡7퐵푎푑푠 − 퐸1 푒푡ℎ푦푙푒푛푒,𝑔푎푠.
Further statistical analysis is performed through use of the Boltzmann probability for i-th
−퐸 /푘 푇 configuration (Pi) by taking the Boltzmann distribution of each minimum (푒 푖 퐵 ) divided by the sum of the distributions of all relevant low energy minima:
푒−퐸푖/푘퐵푇 푃 = , 푖 Σ 푒−퐸푖/푘퐵푇 where Ei is the i-th configuration energy of a gas phase cluster cluster (i.e., Eform as defined above), adsorbed cluster (퐸[Surf + Pt푚B푛]) or adsorbed cluster with a reagent (퐸[Surf + Pt푚B푛 +
푟푒푎푔]), 푘퐵 is the Boltzmann constant, and T is the temperature.
The entropic contribution of relevant minima may also be found by considering the fundamental thermodynamic relation of the Helmholtz free energy (F = U – TS). Specifically, the
Gibbs’ entropy equation (SG) allows us to analyze the effect of discrete states with their respective
Boltzmann probabilities on the ensemble of particular cluster types:
푆퐺 = −푘퐵 ∑푖 푃푖 ln(푃푖), where the PI are the Boltzmann weights and 푘퐵 is the Boltzmann constant. In this way, we may analyze the entropic contribution at a catalytically relevant temperature (TSG).
Intracluster bonding was evaluated through a summation of the electrostatic potential present in a cluster:
푞푖푞푗 푉퐶 = 푘푒 ∑푖,푗 , 푟푖,푗 where qi and qj represent two different atoms, ri,j is the distance between them, and ke is Coulomb’s
1 constant . 4휋휀0 159
B.3 Theoretical Discussion
The geometric diversity present in Pt7 is enhanced even further in Pt7B with 10 isomers. This results in a substantial increase in the configurational entropy’s contribution to the free energy of the system (Table B.3-B.4). In a previous study, 1:1 ratios of Pt:Pd were preferentially stabilized and sintering-resistant over other ratios of Pt:Pd due to intra- cluster bonding and the configurational entropy arising from the presence of many isomers.1 Due to the limitation of our cluster sizes, the sintering modality is limited to extrapolation from Pt and B potential energy surfaces. Pt’s PES yields typical behavior of
~2 eV sintering penalty, observed in other systems, but B’s PES resists facile characterization. The B monomer adsorbs nearly as strongly as the B-doped Pt clusters and distorts the surface by abstracting surface O-Al from their initial positions, raising them by
0.2-2.4 Å (Fig. A.6). Unlike other dopants such as the weakly-bound Zn2, which can
1 evaporate from oxides such as TiO2 and MgO, or the mobile Pd/Pt monomers, B will resist sintering by Ostwald ripening (Table B.3). Thus, borated Pt may exhibit considerable stability compared to other dopants due to the stability of the B-Osurf anchor, the entropic influence of many isomers lowering the free energy of the system, and the intracluster attraction present in the clusters.
In order to analyze the effects of both temperature and reagent adsorption on structural reformation, Molecular Dynamics calculations were performed at higher temperatures relevant to catalysis. In a previous publication, we predicted that in pure Pt7 clusters, prismatic geometries would stabilize to single-layer geometries during ethylene adsorption. We observe this in MD trajectories at 450, 700, and 1000 K. At 450 and 700
K, the prismatic geometry undergoes multiple transformations: it opens up from a prism 160 to a distorted hexagon (circa MD step 150) and varies between other prismatic configurations and other single-layer configurations. Notably, at 450 K after 2000 MD steps, the prismatic geometry has transformed to a configuration very similar to the single-layer. At higher temperatures such as 1000 K, the cluster may also fragment into extended, branched configurations that may be a precursor to the mobile monomers that contribute to sintering by Ostwald ripening. Moreover, the sp2 adsorbed ethylene may interconvert to sp3 adsorption. In MD trajectories of ethylene adsorbed on the single- layer Pt7, at 450 and 700 K, the stability of this configuration leads to only twisting and rotation of ethylene at circa 2000 MD steps with little to no change in the Pt cluster’s geometry.
B.4 References
1. Baxter, E. T.; Ha, M.; Cass, A. C.; Alexandrova, A. N.; Anderson, S. L. ACS Cat. 2017, 7, 3322-3335.
2. Desrosiers, R. M.; Greve, D. W.; Gellman, A. J. J. Vac. Sci. Tech. A: Vac. Surf. Films. 1997, 15, 2181-2189.
3. Krawczyk, M.Appl. Surf. Sci. 1998, 135, 209-217.
4. Rodriguez, J. A.; Truong, C. M.; Corneille, J.; Goodman, D. W. J. Phys. Chem. 1992, 96, 334-341.
5. Ha, M.; Dadras, J.; Alexandrova, A. ACS Cat. 2014, 4, 3570-3580.
6. Dadras, J.; Shen, L.; Alexandrova, A. J. Phys. Chem. C. 2015, 119, 6047-6055. 161
Table B.1 Formation energies of global minima of Pt7, Pt8, and Pt7B
Pt7 Pt8 Pt7B Eform (eV) -27.05 -31.97 -35.38 Eform/atom (eV) -3.86 -4.00 -5.05
Table B.2: Gas phase isomers
Cluster Isomer Δ Eform (eV) P450K P1000K I 0.00 59.90% 35.26% II 0.06 14.19% 18.44% III 0.06 12.84% 17.63% Pt7B IV 0.07 9.72% 15.56% V 0.12 2.83% 8.93% VI 0.18 0.52% 4.17%
Table B.3: Characteristics of borated platinum vs pure platinum clusters
Pt7 Pt8 Pt7B Isomers 5 7 10 ΣP450K Eads (eV) -5.08 -4.89 -4.60 ΔQ (e) -1.22 to -1.44 -0.94 to -1.24 -0.03 to -1.17 T450KSG (eV) 0.010 0.002 0.012 T700KSG (eV) 0.034 0.037 0.061 Es—Pt 2.17 Es—B 3.28 162
Table B.4: Adsorbed isomers of Pt7, Pt8, Pt7B with Boltzmann populations (P) at experimentally relevant temperatures of 450 and 1000 K and charge transfer (ΔQ) from the support
Cluster Isomer Δ Eform (eV) P450K P700K ΔQ (e) VC (eV) I 0.00 75.27% 65.89% -1.22 -3.63 II 0.04 24.49% 32.02% -1.44 -1.32 III 0.24 0.16% 1.26% -1.33 -1.35 Pt7 IV 0.27 0.07% 0.77% -1.25 -3.29 V 0.43 <0.01% 0.05% -1.23 -2.79 ΣP Eads (eV) -5.08 -5.07 TSG (eV) 0.010 0.019 I 0.00 97.65% 87.73% -1.24 0.99 II 0.17 1.14% 5.03% -1.07 0.22 III 0.19 0.74% 3.82% -0.94 -1.37 IV 0.21 0.41% 2.61% -1.08 0.66 Pt8 V 0.30 0.04% 0.57% -1.22 0.59 VI 0.39 <0.01% 0.14% -1.08 0.72 VII 0.41 <0.01% 0.10% -0.80 -0.47 ΣP Eads (eV) -4.89 -4.87 TSG (eV) 0.002 0.009 I 0.00 81.22% 59.26% -0.37 -11.85 II 0.10 6.40% 11.57% -0.70 -25.41 III 0.10 6.39% 11.56% -0.44 -10.95 IV 0.12 3.54% 7.91% -0.03 -6.83 V 0.16 1.37% 4.29% -1.17 -4.49 VI 0.19 0.62% 2.59% -0.95 -22.08 Pt7B VII 0.21 0.38% 1.88% -0.96 -30.11 VIII 0.29 0.05% 0.52% -0.64 -18.84 IX 0.31 0.03% 0.34% -0.91 -30.32 X 0.39 <0.01% 0.09% -1.04 -21.92 ΣP Eads (eV) -4.66 -4.78 TSG (eV) 0.013 0.036 163
Figure. B.1: Gas phase isomers of Pt7B at catalytically relevant temperature of 1000K. 164
Figure B.2: Global minima of adsorbed diborane on Pt7, Isomer I (prismatic) and II (single-layer). 165
Figure B.3: The highest affinity configurations from Pt (light gray) and B (light pink) potential energy surface calculations.
166
Figure B.4: The lowest minima of ethylene adsorbed on Pt7Bglob. 167
Figure B.5: The lowest minima of ethylene adsorbed on adsorbed Pt7B, Isomer II.
168
Figure B.6: The lowest minima of ethylene adsorbed on adsorbed Pt B, Isomer V. 7
169
Figure B.7: The lowest minima of 2 ethylene adsorbed on adsorbed Pt7B, Isomer I.
170
Figure B.8: The lowest minima of 3 ethylene adsorbed on adsorbed Pt7B, Isomer I.
171
Figure B.9: Highlights from the MD trajectories of ethylene adsorbed on prismatic Pt at catalytically relevant temperatures. 7
172
Figure B.10: Highlights from the MD trajectories of ethylene adsorbed on single- layer Pt7 at catalytically relevant temperatures.
173
Figure B.11: Highlights from the MD trajectories of ethylene adsorbed on Pt7B, Isomer I at catalytically relevant temperatures.
174
Figure B.12: Highlights from the MD trajectories of ethylene adsorbed on Pt7B, Isomer II at catalytically relevant temperatures.
175
Figure B.13: Highlights from the MD trajectories of ethylene adsorbed on Pt7B, Isomer V at catalytically relevant temperatures.
176
Figure B.14: The C-sticking configurations of the lowest six isomers of Pt7B.
APPENDIX C
SUPPORTING INFORMATION FOR CHAPTER 4
178
C.1 TPD Fitting Method and Results
In order to determine the distribution of desorption energies for the various borane products observed, TPD/R spectra were fit to the second order rate equation:
−퐸 −푑휃 퐼(푡) 훼 = (휃2(퐸) ∙ 휈) 푒푘푇(푡), 푑푡 where I(t) is the desorption as a function of time, ν is a prefactor, and T(t) is the temperature as a function of time. A distribution of population in sites with different energies for desorption, θ(E), is assumed and ran through the simulation, the calculated
I(t) is compared to the experimental desorption versus time, and then θ(E) is adjusted until the simulated I(t) matches the experiment. Because size-selected cluster samples are time consuming to prepare, and irreversibly changed by a single TPD/TPR run, it is simply not practical to extract ν from a series of coverage-dependent experiments on every cluster size. Therefore, the simulations were tested for ν ranging from 1013 to 1015 s-1, covering a range often found in TPD. The simulated desorption/dehydrogenation energies shift by only ~7% per order-of-magnitude variation in ν, and in Figures S2-3 we present the θ(E) distributions obtained for ν = 1014 s.
179
12000
m/q = 11 Pt7/alumina m/q = 26 10000 m/q = 48 m/q = 59
8000
6000
4000
2000
0
Counts Pt-free alumina film 10000
8000
6000
4000
2000
0 100 110 120 130 140 150 160 170 180 190 200 210 220 230 240 250
Temperature (K)
Figure C.1: Raw TPD spectra for select ion masses produced by electron impact ionization of species desorbing from Pt7/alumina, and from a sample of the alumina/Ta(110) support, following exposure to 1.5 L of diborane at 130 K, followed by heating at 3 K/second.
180
Figure C.2: Desorption energy distributions obtained by fitting the m/q = 11 desorption temperature dependences for B2H6 TPD/R experiments on Pt-free alumina and Pt7/alumina.
181
Figure C.3: Desorption energy distributions obtained by fitting the m/q = 26, 48, 59 desorption temperature dependences for B2H6 TPD/R experiments on Pt-free alumina and Pt7/alumina.
182
1600 As-deposited Pt 1400 7 1.5 L B2H6 - 130 K 1200
1000
800
Counts 600
400
200
0 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
E/E
Figure C.4: Raw ISS for Pt7/alumina samples. A spectrum measured immediately after depositing 0.1 ML of Pt7 is shown in blue. A spectrum taken immediately after exposing a Pt7/alumina sample to 1.5 L B2H6 at 130 K is shown in red. Note the large attenuation of Pt ISS signal (E/E0 ≈ 0.93), and smaller attenuations of the Al and O signals (E/E0 ≈ 0.6 and 0.41). Note also that no significant growth of signal is seen in the region expected for boron (E/E0 ≈ 0.26) after diborane exposure.
183
0.0020 D 0.0018 2
0.0016
0.0014
0.0012
atom/sec
0.0010
0.0008
0.0006
moleucles/pt
2
D 0.0004
0.0002
0.0000 0 100 200 300 400 500 600 700 800
Temperature (K)
Figure C.5: TPD of D2 from Pt7/alumina, dosed with 5 L of D2 at 130 K, then heated at 3 K/sec.
184
Computational
Figure C.6: Gas phase diborane with an Eform of -33.90 eV.
185
Figure C.7: Local minima of adsorbed diborane on Pt7, Isomer I (prismatic).
186
Figure C.8: Local minima of adsorbed diborane on Pt7, II (single-layer).
187
Figure C.9: Local minima of gas phase Pt4B4.